![24
efficiency that focussed on technical efficiency alone would conclude that both
the providers are equally efficient. This would contrast with the outcome of a
study on economic efficiency for the following reasons. If the technology and
input mix were similar between providers, the provider facing higher input prices
would exhibit the lower economic efficiency. Furthermore, if factor prices vary
among providers, no inferences can be drawn about economic efficiency, unless
a more careful analysis is conducted. Social cost of resources, instead of factor
prices, may have to be considered in such cases.
There are numerous studies which examined some of the above mentioned
issues in detail. These include, Bays (1980), Pauly (1978), Baron (1978), Elnicki
(1974), Lee Wallace (1973), Lave, Lave and Silverman (1972), Hu (1971),
Francisco (1970), Cohen (1967), Cowing and Holtman (1983), Watts and
Klastonin (1980), Luke (1979), Feldstein and Shutting (1977), Bays (1977),
Thompson, Fetter and Mross (1975), Rafferty (1971), and Berry (1970).
2.1.3. Studies for developed countries
During the past two and half decades, estimation of cost function for hospitals
has become a common place for industrialised countries. Reviews by Cowing,
Holtman and Powers (1983), Wagstaff (1989), and Ricardo (1992) have
documented this literature. The initial attempts to estimate cost functions using
data from hospitals in industrialised countries employed composite measures of
hospital output (e.g. Cohen 1967). They used average or unit cost of inpatient
day or admission as the dependent variable, and a variety of interrelated
explanatory variables such as occupancy rates, patient flow, length of stay, and
capacity as explanatory variables (Mann and Yett 1968).
Most of the early empirical studies [Feldstein (1967), Berry (1967), Carr and
Feldstein (1967), Ingbar and Tayler (1968), Fransisco (1970)] were concerned
with the effect of size on hospital costs – that is, the existence of scale effects
and the related questions of optimal size. Authors of these studies estimated
rather restricted versions of the general cost function, usually without recognizing
many of the implicit assumptions involved. For example, most of the models did
not include input prices in the estimated cost function, an omission which is
equivalent to assuming zero input substitutability, that is, a fixed proportions
technology for producing hospital care.](https://image.slidesharecdn.com/87a070eb-f656-41af-b821-bb885c0e754d-160602230301/85/PHD-THESIS-ENTIRE-DOCUMENT-39-320.jpg)
![33
countries. To the best of our knowledge, there are no published studies in India
using cost function approach for measuring the hospital performance and
efficiency. The present study aims at filling this gap by using data on secondary
level district hospitals in AP.
2.2. Financing of health services
Government intervention in the provision of health care services is supported on
the grounds of externality, equity and efficiency. Several economists like Arrow
(1963), Newhouse (1970), and Cuyler (1971) gave sufficient justification for the
government provision of health services. They argue that private providers can
not supply the health services due to market failure. Since health services are
basic and they help in improving the quality of human capital and thereby
economic development, there is a need for the government provision of these
services. World development report (1993) reiterates this view. Further, the
report states that the government provision of health services helps in reducing
poverty, inequity and inefficiency.
As already discussed in Chapter I, World Health Organizations’ Alma-Ata
declaration of 1978 outlined a global strategy for ‘health for all by 2000’ through
the primary health care system. WHO estimated that the implementation of this
strategy requires annual per capita health expenditure of US $15 in 1981 prices
for most of the developing countries. Since per capita public spending on health
is currently only US $2.3, there is an annual resource gap of US $50 billion for all
the developing countries. Even if developing countries could fund as much as
50% of this amount, they would have to seek external funding of about seven
times the present level of international transfers. The growth of per capita
domestic public spending on health, therefore, is not anywhere near the levels
required to meet the goals of the global strategy. [WHO (1981c), World Bank
(1980b)].
Because of this reason and others (discussed in Chapter VI), various
researchers and policy makers suggest alternative method of generating funds to
finance health services (Jimanez 1986a, 1986b). Among the suggested
measures are; user fees, development of suitable insurance system etc. It is in
this context that the present study aims at examining the issue of financing the
health services in developing countries in general and AP in particular. The study
aims at reviewing the traditional arguments in favor of government intervention](https://image.slidesharecdn.com/87a070eb-f656-41af-b821-bb885c0e754d-160602230301/85/PHD-THESIS-ENTIRE-DOCUMENT-48-320.jpg)
![47
k
1j
jj98 βpRˆlnαBlnαNln
k
1j
jj1110 βpRˆlnαBlnαHln
Feldstein (1967) compared the results of this recursive model with a similar set of
single equation regressions, also in Cobb-Douglas form. The comparison favored
the recursive model. The results also indicate, rather weakly, the presence of
decreasing returns to scale in hospital services.
Cost functions
For a variety of reasons, it is easier to study the aspect returns to scale using
cost functions (Cowing, Holtman and Powers, 1983) and this in any case, is the
way most of the studies in health economics have approached this topic. In
addition, many hospital studies are primarily concerned with cost behavior rather
than input-output relationships addressed directly by production functions. A
simple cost function may be represented as:
p)(Y;CC (3.11)
where total cost (C) is a function of the output vector Y and the vector of input
prices (p). The cost function has the ability to consider the effect of multiple
outputs on total cost. On the other hand, it deals with inputs only implicitly
although, provided technical efficiency can be assumed, modern theories of
Shephard-Samuelson duality permit one to move between cost and production
functions to remedy this deficiency. In any case, ‘p’ serves as a vector of
parameters with (under optimising behavior) the vector of outputs (Y) resulting
from input mixes and magnitudes selected under whatever technology (Baumol,
Panzar and Willig 1982).
Shephard [reproduced in Baumol , Panzar and Willig (1982)] used equation
(3.11) for his duality relations between cost and production functions, which inter
alia makes it possible to move in a relatively easy manner between cost and
production functions. Shephard lemma (see for example, Bothwell and Cooley
1978) states:](https://image.slidesharecdn.com/87a070eb-f656-41af-b821-bb885c0e754d-160602230301/85/PHD-THESIS-ENTIRE-DOCUMENT-62-320.jpg)
PHD THESIS - ENTIRE DOCUMENT
- 1. I EFFICIENCY AND FINANCING OF HEALTH SECTOR - A METHODOLOGICAL DEBATE FOR SECONDARY LEVEL HOSPITALS IN ANDHRA PRADESH, INDIA BY DR PURNA CHANDRA DASH1
- 2. II 1About the book and the author: A Ph.D. in Economics, with specialization in Monitoring and Evaluation, and Management and Coordination in the fields of Health Financing, Provider Payment Mechanism (PPM),Health Sector Efficiency (econometric and other approaches), Health sector and Hospital Costing (preparation of MTEF and Public Expenditure Review), Burden of diseases and Cost Effectiveness Analysis, Health System Strengthening through Management Interventions(DHS),Financial Management (Planning, Budgeting, Monitoring and budget tracking), National Health Accounts and Public Private Partnership(PPP)in Health, with working experience of over 15yrs in health programs, health systems, primary health care & international health. Experience of managing health programs of several Million USDs nationally and internationally as specialist, manger or adviser, for several agencies/organizations receiving funds from different donors, experience of working in several countries including those in Asia, Africa and Europe. Good at identifying and promoting new approaches (Health Equity Fund, Services Delivery Grants )on health financing and other relevant aspects of health care delivery in a given setting. Having extensive experience in networking and negotiations and having a passion for bringing affordable health for all. He is presently heading the Health Unit of Centre for management and Social Research – Health Division, Jaipur, India. The present manuscript is his Doctoral Thesis which was submitted to University of Hyderabad in 1999 under the supervision of Prof. K. N. Mutry and was evaluated by Prof. Peter Berman, Harvard School of Public Health; Prof. Brijesh Purohit, Ex- Professor, Administrative Staff College of India Hyderabad.
- 3. III ACKNOWLEDGEMENT I express my deep sense of gratitude and profound respect to my supervisor Dr. K.N. Murty who has helped and encouraged me at all stages of my thesis work with great patience and immense care. I am particularly indebted to Prof. D. Narasimha Reddy for his helpful comments on the earlier draft of the thesis and his constant encouragement to participate in all spheres academic activities. I wish to express my gratitude to Dr. V. Narasimhan, Senior Lecturer, Department of Economics, for providing useful insights on different aspects of health financing. I am also thankful to Dr. A. V. Raja, Reader, Department of Economics, for spending his valuable time on discussing various microeconomic aspects related to my work. Also I have to mention Dr. S. Sandhya, Research Scientist, Department of Economics for her interest and useful discussions about my work through out. My sincere thanks are to Prof. V.V.N. Somayajulu, Dean, School of Social Sciences, Prof. G. Nancharaiah, Head, Department of Economics, University of Hyderabad and Prof. M. Atchi Reddy for their support and help in completing my work successfully. I am deeply indebted to Dr. Sachidananda Mohanty, Reader, Department of English for his invaluable help and co-operation during my stay in the university campus. His brotherly attention towards me has influenced my academic as well as non-academic career to a large extent. I wish to express my gratefulness to Dr. Prasanta Mahapatra, President, Dr. G. N. V. Ramana, Former Director, Institute of Health Systems, Hyderabad and Dr. Ramesh Durvasula, Chairman, Social Services Area, Administrative Staff College of India, Hyderabad; who introduced me to the exciting discipline of health economics. I am very much indebted to my dearest friend, Mr. Abdullah Al-Hamzi for his encouragement, timely help and invaluable support during the course of my Ph.D. work. I unfailingly remember the friendship of Lalat for his warm and affectionate feeling and keen interest in my work. I am equally thankful to my
- 4. IV friends Mohd. Ali, Pramod Bhai and Debasis Bhai for their company and wonderful time we shared together in the campus of Hyderabad University. I am grateful to Mr. Nagarjuna, Project Director, Andhra Pradesh first referral health systems project; Dr. Saila Quraishi, Commissioner, Andhra Pradesh Vaidya Vidhana Parishad for providing me necessary data and information to carry out the study. The invaluable help provided by Mr. Joseph and the staff of computer centre, APVVP, for clarifying doubts about the data is duly acknowledged. I am thankful to staff of Centre for Economic and Social Studies, Hyderabad; Institute for Social and Economic Change, Bangalore. The help and co-operation provided by Dr. D. T. Jamison, World Bank; Prof. P. Berman and Prof. C.J. Murry, Harvard School of Public Health is duly acknowledged. I would like to express my special thanks to my classmates, Mr. G. Laxminarayana and Mr. Hemant Kumar, for their encouragement and timely help. I do not know how to adequately thank Fanta, Sujeet, Surath, Sanatan, Mukund, Anjan, Chandan, Ghanashyam and Pratap Jena for the sincere friendship and affectionate relationship we had together during my stay in Hyderabad. I am really indebted to my friends, Ramakanta, Ratikanta, Ranjan, Saroj, Niranjan, Umakanta, Sadananda, Purna (Parida and Padhan), Mirza Allim, Pratap, Prabir, Manas, Bama, Panchanan, Mahanand, Bhagirath, Jayaprakash, Hrushi and others who are directly or indirectly helped me during my stay in the campus. Finally, but not the least, I owe sincere respect to my mother Kanchanabala Dash, whose encouragement, love and affection made me what I am. I am thankful to my sisters Kalpana, Bandana, and my little brother Kishore (Tukula) for their continuous encouragement, understanding and unlimited support throughout my studies.
- 5. V LIST OF ABBREVIATIONS AC ALS APVVP BOR BTR CBR CDR CES CHV CUP CV DHA DHS DRG EOS FTE FW ICD : Average Cost : Average Length of Stay : Andhra Pradesh Vaidya Vidhana Parishad : Bed Occupancy Rate : Bed Turnover Rate : Crude Birth Rate : Crude Death Rate : Constant Elasticity of Substitution : Community Health Volunteers : Combined Utilisation and Productivity : Coefficient of Variation : District Health Authority : Directorate of Health Services : Diagnostic Related Groupings : Economics of Scale : Full Time Equivalent : Family Welfare : International Classification of Disease
- 6. VI IHO IMR LER MAPE MC MCH MNP MPW NHP OLS PHC RMPE ST TC VES WHO WLS : Intermediate Health Unit : Infant Mortality Rate : Life Expectancy Rate : Mean Absolute Percentage Error : Marginal Cost : Maternity and Child Health : Minimum Needs Programme : Multi Purpose Workers : National Health Policy : Ordinary Least Squares : Primary Health Center : Root Mean Percentage Error : Scheduled Tribe : Total Cost : Variable Elasticity of Substitution : World Health Organization : Weighted Least Squares LIST OF TABLES
- 7. VII Title Page Table 1.1: General health statistics, 1944 5 Table 1.2: Health indicators of India, 1961-96 10 Table 1.3: Life expectancy at birth (yrs) by sex and residence 10 Table 1.4: Health indicators of Andhra Pradesh and India 12 Table.1.5: Socio economic indicators of Andhra Pradesh and India 13 Table 4.1: Number of hospitals for which data are available 57 Table 4.2: Summary statistics on bed capacity of APVVP hospitals during 1996 58 Table 4.3: Summary statistics on hospital manpower in 1996 59 Table 4.4: Summary statistics of hospital performance indicators for district Hospitals 61 Table 4.5: CUP Sector wise distribution of district level hospitals 65 Table 4.6: Summary statistics of hospital performance indicators for area Hospitals 68 Table 4.7: CUP Sector wise distribution of area level hospitals 69 Table 4.8. Summary statistics of hospital performance indicators for community Hospitals 70 Table 4.9: CUP Sector wise distribution of community level hospitals 72 Table 4.10: Summary statistics of the variables in BOR and BTR regressions 76 Appendix Table 4.1: Results using total number of doctors, nurses, paramedical, and other staff as independent variables 83 Table 5.1: Summary statistics of hospital costs, outputs, and inputs for district hospitals in AP. 93 Table 5.2: Parameter estimates and related measures of simple total cost function for district hospitals in AP. 96 Table 5.3: Observed and estimated factor input levels (No.) using simple cost function for district hospitals in AP, 1995 97-98 Table 5.4: Estimated marginal cost (Rs) for each output category using simple cost function for district hospitals in AP. 103 Table 5.5: Parameter estimates and related measures of hybrid cost function for district hospitals in AP. 106 Table 5.6: Observed and estimated factor input shares (percent) using hybrid cost function for district hospitals in AP, 1995 107- 108
- 8. VIII Table 5.7: Parameter estimates and related measures of Cobb-Douglas cost function for district hospitals in AP. 116 Table 5.8: Parameter estimates and related measures of logarithmic quadratic cost function for district hospitals in AP. 116- 117 Table 5.9: Parameter estimates and related measures of translog cost function for district hospitals in AP. 118- 119 Table 5.10: Estimated R2, MAPE and RMPE values for different models. 120 Table 5.11: Estimated log likelihood and 2 statistic values 122 Table 5.12: Observed and estimated factor input shares (percent) using translog cost function for district hospitals in AP,1995 123- 124 Table 5.13: Observed and estimated factor input levels (No.) using translog cost function for district hospitals in AP, 1995 126- 127 Table 5.14: Observed and estimated factor ratios with respect to fixed factor (beds) using translog cost function for district hospitals in AP, 1995. 130- 131 Table 5.15: Estimated product-specific cost elasticities and overall economies scale using translog cost function for district hospitals in AP. 132 Table 5.16: Estimated economies of scope between pairs of outputs using translog cost function for district hospitals in AP. 134 Table 5.17: Estimated Allen-Uzawa partial own elasticities of factor substitution using cost function for the district hospitals in AP. 136 Table 5.18: Average Allen-Uzawa partial elasticities of factor substitution using translog cost function for the district hospitals in AP. 138 Table 5.19: Estimated own-price elasticities of factor demand using translog cost function the district hospitals in AP. 139 Table 5.20: Average elasticities of factor demand using translog cost function for the hospitals in AP. 140 Appendix 5A.1: Basic data on annual salary (Rs) for district hospitals in AP. 142 Appendix 5A.2: Basic data on inputs and outputs of district hospitals in AP. 143 Appendix 5B: Estimation of conversion factor between IP and OP: 144- 145 Appendix 5C.1: Observed and estimated factor input shares (percent) for district level hospitals in AP, 1995. 146- 147 Appendix 5C.2: Observed and estimated factor input shares (percent) for district level in AP, 1996. 148- 149 Appendix 5C.3: Observed and estimated factor input levels (No.) for 150-
- 9. IX district level hospitals in AP, 1995. 151 Appendix 5C.4: Observed and estimated factor input levels (No.) for district level hospitals in AP, 1996. 152- 153 Appendix 5C.5: Estimated product specific cost elasticities and overall economies of scale for all district hospitals in AP, 1995. 154 Appendix 5C.6: Estimated product specific cost elasticities and overall economies of for all district hospitals in AP, 1996. 155 Appendix 5C.7: Estimated economies of scope by pairs of output for all district hospitals AP, 1995. 156 Appendix 5C.8: Estimated economies of scope by pairs of output for all district hospitals AP, 1995. 157 Table 6.1: Share of government health expenditure (percent) in different countries 164 Table 6.2: Real government expenditure on health and health related services Andhra Pradesh 173 Table 6.3: Age and sex distribution of sample hospital users, Karimnagar 178 Table 6.4: Educational status of sample hospital users, Karimnagar 179 Table 6.5: Occupational distribution of sample hospital users, Karimnagar 181 Table 6.6: Distribution of sample users of Karimnagar hospital based on main earner's Occupation 181 Table 6.7: Distribution of sample users of Karimnagar hospital based on income 182 Appendix Table 6A.1: Real government expenditure (Rs. mil) on health and health services in Andhra Pradesh 187 Appendix Table 6A.2: Distribution of real expenditure (Rs. mil) on public health and hospitals 188 Appendix Table 6A.3: Composition of user fees (Rs.) collected from APVVP district hospitals during 1995 189 Appendix Table 6A.4: Composition of user fees (Rs.) collected from APVVP district during 1996 184 Appendix 6B: Questionnaire used for survey of hospital users in Karimnagar district, Andhra Pradesh 191- 196
- 10. X LIST OF CHARTS Title Page Chart 4.1: Sectoral location of district level hospitals based on average BOR and BTR values for 1991-96 64 Chart 4.2. District level hospitals (%) lying in different sectors over time 66 Chart 4.3: Sectoral location of area hospitals based on average BOR and BTR values for 1991-96 68 Chart 4.4. Area level hospitals (%) lying in different sectors over time 70 Chart 4.5: Sectoral location of community hospitals based on BOR and BTR values for 1991-96 72 Chart 4.6. Community level hospitals (%) lying in different sectors over time 73 Appendix Charts 4.1 to 4.3: Sectoral location of district level hospitals based on BOR and BTR Values 84 Appendix Charts 4.4 to 4.6: Sectoral location of district level hospitals based on BOR and BTR Values 85 Appendix Charts 4.7 to 4.9: Sectoral location of area level hospitals based on BOR and BTR Values 86 Appendix Charts 4.10 to 4.12: Sectoral location of area level hospitals based on and BTR Values 87 Appendix Charts 4.13 to 4.15: Sectoral location of community level hospitals on BOR and BTR Values 88 Appendix Charts 4.16 to 4.18: Sectoral location of community level hospitals based on BOR and BTR Values 89 Graph 5.1: Estimated total cost curve of IP days using simple cost function for secondary level district hospitals in AP. 101 Graph 5.2: Estimated average and marginal cost curves of IP days for secondary level district hospitals in AP using simple cost function. 102 Chart 6.1: Real government expenditure (Rs mil) on health and related services,1983-96 172 Chart 6.2: Real government health expenditure (Rs mil) on different components, 1983- 96 174 Chart 6.3: Real government expenditure (Rs mil) on public health services,1983-96 175 Chart 6.4: Real government expenditure (Rs mil) in secondary and tertiary level hospitals, 1983-96 176
- 11. XI CONTENTS Acknowledgement List of abbreviations List of tables List of charts Page CHAPTER I: HEALTH SECTOR IN INDIA: AN OVERVIEW 1- 17 1.0. Introduction 1.1. Definition of health 1.1.1. Health status and health care 1.2. Health status in India 1.2.1. Pre-Independence period 1.2.2. Post-Independence period (1951–1997) 1.2.3. Present health status 1.3. Health status in Andhra Pradesh 1.4. Allocation of funds to health sector in AP 1.5. Motivation for the present study 1.6. Broad objectives of the study 1.7. Chapter outline 1 2 3 3 3 4 9 11 15 16 16 16 CHAPTER II: REVIEW OF LITERATURE 18-35 2.0. Introduction 2.1. Studies on hospital efficiency 2.1.1. Studies using performance indicators 2.1.2. Econometric studies 2.1.3. Studies for developed countries 2.1.4. Studies for developing countries 2.1.5. Studies for Andhra Pradesh 2.2. Financing of health services 2.3. Conclusion 18 18 18 18 25 28 33 33 34 CHAPTER III: HOSPITAL PERFORMANCE AND EFFICIENCY ALTERNATIVE APPROACHES: 36-54
- 12. XII 3.0. Introduction 3.1. Cost accounting method 3.1.1. Step down procedure 3.1.2. Other accounting methods 3.2. Ratio analysis 3.3. Combined utilization and productivity analysis 3.4. Statistical/econometric approaches 3.4.1. Technical and economic efficiency 3.4.2. Evaluation of hospital efficiency 3.4.3. Production and cost functions 3.5. Specific objectives of the study 36 36 36 37 38 39 41 41 43 44 54 CHAPTER IV: HOSPITAL PERFORMANCE: AN EVALUATION 55- 89 4.0. Introduction 4.1. Sampling and Methodology 4.1.1. Sampling 4.1.2. Methodology 4.2. Empirical results 4.2.1. District hospitals 4.2.2. Area hospitals 4.2.3. Community hospitals 4.3. Determinants of hospital performance indicators 4.3.1. Model 4.3.2. Data 4.3.3. Empirical results 4.4. Summary and conclusion 55 55 55 58 60 60 65 69 73 73 74 75 80 CHAPTER V: HOSPITAL EFFICIENCY: AN ANALYSIS 90- 157 5.0. Introduction 5.1. Data on hospital costs and outputs 5.2. Simple short-run cost function 5.3. Flexible hybrid cost function 5.4 Structural cost functions 5.5 Summary and conclusion 90 90 90 94 104 109
- 13. XIII CHAPTER VI: FINANCING HEALTH SERVICES: AN ASSESMENT 158- 196 6.0. Introduction 6.1. Health care activities 6.2. Characteristics of health care 6.3. Intervention in preventive care 6.4. Intervention in curative services 6.5. Financing health interventions 6.6. The Indian case 6.6.1. Pattern of health expenditure in AP 6.6.2. User fees and its present status 6.7. Case study 6.7.1. Field survey 6.7.2. Summary 6.8. Conclusion 158 159 159 160 162 165 170 171 176 177 177 184 185 CHAPTER VII: SUMMARY AND CONCLUSION 197- 203 BIBLIOGRAPHY 204-214
- 14. XIV DECLARARTION This is to state that the thesis entitled “Efficiency and Financing of Health Sector: A Study of Secondary Level Health Services in Andhra Pradesh” submitted to the Department of Economics, University of Hyderabad for the award of degree of Doctor of Philosophy in Economics, is the original work carried out by me under the supervision of Dr. K. N. Murty, Reader, Department of Economics, University of Hyderabad; and the same has not previously formed the basis for the award of any degree, diploma, associateship, fellowship or any other similar title of recognition. Hyderabad Purna Chandra Dash Date:
- 15. XV Dr. K. N. Murty Reader Department of Economics School of Social Sciences University of Hyderabad CERTIFICATE I certify that the thesis entitled “Efficiency and Financing of Health Sector: A Study of Secondary Level Health Services in Andhra Pradesh” submitted by Mr. Purna Chandra Dash, to fulfil the requirement for the award of the degree of Doctor of Philosophy in Economics, has been carried out under my supervision and no part of the thesis has been submitted for any degree or diploma of any other university. Hyderabad Dr. K. N. Murty Date: (Supervisor) Head of the Department Dean of the School
- 16. 1 CHAPTER - I HEALTH SECTOR IN INDIA: AN OVERVIEW 1.0. Introduction Health is a fundamental human right and a worldwide social goal. Health is necessary for the realization of basic human needs and to attain the status of a better quality of life. In 1977, the 30th World Health Assembly decided that the main social target of governments and World Health Organization (WHO) in the coming decades should be "…the attainment by all the citizens of the world by the year 2000 of a level of health that will permit them to lead a socially and economically productive life" (WHO 1979). Such a declaration has led most of the governments in western countries to give much more priority to their health care systems through higher allocation and better utilization of resources in order to improve the quality of health care. Less developed countries are in the process of improving it and some among them are yet to start. India also has been attempting towards this end. The major hindrances on its way could be attributed to inadequate allocation of resources for the health sector, rapid population growth, inefficient use of the resources allocated and above all lack of public consciousness about their own health status. Health being a State subject in the Indian federal system, different states in the country have been trying to meet the WHO health goal through mobilization of both internal and external resources including the funds from foreign agencies. Specifically, the state of Andhra Pradesh has been in the forefront in this regard and somewhat successful in developing a better public health care delivery system. However, the achievement of the goal of "health for all" for the state is perhaps still a distant dream. Here a major point that needs to be understood is that the country needs to give emphasis on the rural health services where nearly 70% of total Indian population still lives. Despite repeated pronouncements by the policy makers about the need for rural emphasis, health services provided to the people have continued to be urban oriented where a major chunk of the resources allocated to the health sector are spent. In this chapter we attempt to give an outline of the functioning of the health care delivery system of India in general and Andhra Pradesh in particular. Before going into detailed debate on the issues involved it may be useful to clarify certain basic
- 17. 2 concepts that are frequently used in health care research. 1.1. Definition of health Different professional groups define the concept of health in different ways. Medical professionals define health in terms of illness, which, in turn is expressed in terms of mental or physical disorders. This concept of health is predominantly based on pathology and is concerned with the presence or absence of disease and the stage of its invasiveness. Some others define illness through giving emphasis to the amount of pain suffered or the degree to which individuals are restricted in undertaking their normal day to day activities. For some, the maintenance of health is also linked to social aspects such as unemployment and wealth. The broadest definition of health appears to accept anything and everything that can affect health status. The most widely accepted definition of health given by World Health Organization (WHO 1948), quoted in FRCH (1987) is as follows: "Health is a state of complete physical, mental and social well being and not merely an absence of disease or infirmity". Having defined health in the above fashion, WHO developed the following indicators of health: (a) Health policy indicators: Political commitment to health for all, resource allocation, the degree of equity of distribution of health services, community involvement, and organizational framework and management process. (b) Social and economic indicators: Rate of population increase, GNP or GDP, income distribution, work conditions, adult literacy rate, housing and food availability. (c) Health care indicators: Availability, accessibility, utilization and quality of care. (d) Health status indicators: Birth weight, nutritional status and psychological development of children, infant mortality rate, child mortality rate, life expectancy at birth, maternal mortality rate, disease specific mortality, morbidity incidence and prevalence, disability prevalence (WHO 1981b). Given these indicators of health, the achievement of the goal `health for all' is not only the responsibility of the individuals seeking health care, but also of the
- 18. 3 community, state and international agencies as well. The role of the individual differs from country to country depending upon the social, economic, cultural and environmental conditions. The indicators given above bring out the multidimensional nature of health. All the above indicators are interlinked and the improvement of health status and health care is emphasized in all of them (World Development Report, 1993). 1 1.1.1. Health status and health care These two concepts differ in their basic nature. Health status has a value in use, but contains no value in exchange. A detailed discussion about the measurement of health status such as disease specific mortality, morbidity, incidence and prevalence of different diseases and disability caused by different diseases requires a separate study and is not within our scope. This study aims at establishing a relationship between cost of health care and health output. The latter includes outpatient care, inpatient care, emergency care etc., which are provided through different health care institutions at various levels. This study however focuses on secondary (curative) health care institutions only. As a background for our study, in the next section, we attempt to give a brief historical overview of the health status indicators for India and Andhra Pradesh. 1.2. Health status in India 1.2.1. Pre-Independence period The earliest indigenous system of medicine can be traced back to the development of Vedic medicine following the Aryan migration to the Indus Valley. “The Vedic Samhitas which were religious texts, contain the concepts of anatomy, physiology and pathology which were quite impressive” (Banerji 1974). Ashok Maurya (279 - 236 B.C) was responsible for the spread of social medicine, manifested in a public health care system that included hospitals and Sanitaria for men, women and children (FRCH 1987). 1 Using these indicators as measures of health status, it is argued that, in case of underdeveloped countries health status is low compared to developed countries. The reasons cited are, low income level and higher population growth, which characterize developing countries. Also growing industrialization is another important reason. Industrialization is a major cause of environmental pollution, which in turn leads to chronic diseases.
- 19. 4 During the medieval period, physicians from west Asia who were trained in Unani system, compiled and translated Ayurvedic texts. Hakim Yooufi (16th century) a physician in the court of Baber and Humayun is said to have synthesized Arabian, Persian and Ayurvedic thought and produced a composite and integrated medical system (Leslie 1974). The late 19th and 20th century saw the method of Ayurvedic Physicians who “…were radically different from the classical texts and were deeply influenced by the Galenic (Unani) traditions in Islamic Medicine” (FRCH 1987). Western system of medicine was introduced in India in the latter half of the 18th century mainly to serve the needs of the colonial settlers and their armed forces that came to rule India.2 These facilities were extended to a small segment of people, mostly elite masses, living in urban areas. Those people could get better western medical facilities including health-related services such as sanitation and water supply. A larger segment of people who were staying in the rural areas was neglected except in periods of incidence of massive epidemic diseases such as plague, cholera and small pox. Therefore, the rural population relied upon the indigenous medical system. The report of the health survey and planning committee (Bhore Committee) submitted to the Government in 1946, gave the following health status of the country on the eve of Indian Independence (Table 1.1). Also the committee pointed out that "…high morbidity and mortality were largely preventable and were mainly due to the absence of safe drinking water, sanitation, waste disposal, poor nutrition, lack of preventive and curative health services and poor health consciousness among the people themselves" (Govt. of India, 1946). In addition, the committee pointed out the severe shortage of medical personnel. Independent India inherited a limited infrastructure of modern health services from the British. The medical training was oriented towards the needs of the army medical personnel. It was expanded and institutionalized in the licentiate course offered by the Universities of Madras, Calcutta and Bombay. Looking at the 2 The emergence of an organized public health system, in fact, dates back to the appointment of an inquiry committee, `Royal Commission' in 1859 in order to investigate the causes of poor physical condition in the British Indian army.
- 20. 5 circumstances Bhore committee recommended that at least 15% of total public expenditure should be devoted towards health sector. 1.2.2. Post-Independence period (1951–1997) In contrast to the recommendations of Bhore Committee (1946), it could be seen from various plan documents that the expenditure on health sector has been declining during post-independence period. The outlay has decreased from 3.3% in the first plan (1950-56) to 3% in the second plan (1956-61), 2.6% in the third plan (1961-66), 2.1% in the fourth plan (1969-74), 1.9% in the fifth plan (1974- 79), 1.8% in the sixth plan (1980-85) and seventh plan (1985-90) (Health Statistics of India, 1984) Table 1.1: General health statistics in British India*, 1944 * Figures do not include the princely states. Source: Health Status of Indian People, FRCH, 1987. However, with 3.3% of the total allocation for health sector, the first plan (1951- 56) aimed at: providing safe drinking water to the people, controlling the communicable diseases like malaria, cholera, tuberculosis etc., opening smaller Statistic Value General death rate per 1000 22.4 Infant mortality rate /1000 162 Life expectancy at birth (Years): Females : Males 26.5 26.9 Death rate of children under 10 years /1000 4 Disease Incidence (%) Cholera 2.4 Small Pox 1.1 Plague 0.5 Fever Including Malaria 58.4 Dysentery 4.2 Respiratory diseases 7.6 Others 25.8 Total 100
- 21. 6 health units for providing health services to the rural masses and providing basic family planning services. Since mere opening of health units does not mean better provision of services, the plan also aimed at equipping the health units with trained personnel along with adequate provision of drugs and equipment. One spectacular achievement of the first plan was launching a program to control communicable diseases like malaria, small pox, tuberculosis, leprosy, filariasis, trachoma and cholera. In addition, 725 smaller health units, known as primary health centers (PHC) were also opened during the first plan. The All India Institute of Medical Sciences (AIIMS), various research centers to study the major diseases and a few training centers for paramedical workers were also established. Though various programs were launched and various institutions were opened during the first plan, the benefits of such programs could not be realized during first plan itself. Therefore, the second plan (1956-61) re-emphasized on the same set of objectives with special attention on the implementation of the already existing programs. The Health Survey and Planning Committee (1961a) under the chairmanship of Sir Laksmanswami Mudaliar made an evaluation of the progress in health conditions and health services from the period after the submission of Bhore Committee (1946) report. The committee pointed out that there was a significant progress in controlling the communicable diseases and establishing various institutional facilities during first two five-year plans. The deaths due to cholera, malaria and small-pox were found to be halved and institutional facilities in the form of hospitals and dispensaries had increased from 9068 in 1950 to 13875 in 1960 and the number of hospital beds increased from 112,000 to 200,000 during the same period. The report, however, revealed that "…almost one third of the total number of beds are to be found in 125 teaching hospitals, 47% of the total number of beds are provided by 6.6% of the hospitals with 200 beds and above...". The vast majority of these institutions were situated in urban areas. "Hospitals with a bed strength of less than 50 form over 70% of the total number of hospitals and only 19% of the total beds are to be found in them". The committee admitted that the basic health facilities had not reached at least half the nation. In addition, the committee also reported the non-availability of adequate manpower and infrastructure facilities in the already established health care institutions. Therefore, the committee recommended consolidation, rather
- 22. 7 than expansion, of the existing PHCs and their phased upgrading; achieving the norm of 40,000 population per PHC instead of 60,000; discontinuation of the establishment of new PHCs in the absence of full employment of staff, and equipping of district hospitals with "Mobile Clinics" for the non-PHC population. In view of the drawbacks of first and second five year plans, as pointed out by Mudaliar Committee (1961a), the third and fourth five year plans (1961-66, 1969- 74) gave more emphasis on the training of health manpower that were required for smooth running of already established institutions. The third plan aimed at providing “increased facilities” for linking health and social education. The fourth plan with regard to the PHC program, took up the immediate objective of establishing "an effective machinery for the speedy construction of building, staff, drugs and equipment" which was to become the minimum needs program (MNP). It emphasized on both curative as well as preventive services in rural areas as a base for the control and eradication of communicable diseases. Towards the end of fourth plan, in February 1973, in addition to the preventive and promotive health services, the government announced its intention of providing "specialized services" in medicine, surgery, obstetrics and gynecology. For the first time, substantial amounts were allocated for the training of paramedical cadre (Rs.12.93 Crores). Further, 10 more medical colleges were opened during plan period, bringing anticipated doctor population ratio to 1:1510 in 1968 compared to 1:6100 in 1961. This is in sharp contrast to the rural situation where each PHC with maximum 2 doctors served a population of nearly 120,000. In spite of various attempts made during third and fourth plan in order to reduce regional disparities, the provision of health services were found to be urban oriented. Therefore, the fifth plan (1974-79) aimed at correcting these problems through providing better accessibility to the health services to the rural masses. In addition, the plan also emphasized on developing referral services, improving the quality of medical education and training of health professionals, and above all, integrating the health and family welfare components in order to provide minimum public health services in the country. The chief instruments for achieving these goals were the MNP, the multi purpose worker (MPW) training scheme, and the priority treatment of backward and tribal areas. The MNP which was a package of inputs in the field of health, nutrition, environmental improvement and water supply, elementary adult education, roads and
- 23. 8 electrification in rural areas and housing for land-less laborers received a large outlay of Rs.2,291 crores and the MPW training scheme received Rs.8 crores. In addition, the plan targeted the opening of 101 PHCs, 11,036 sub-centers and 1,293 rural hospitals for rural and remote areas. Further, in 1977 the GOI introduced a major innovation in the nation's health strategy by launching a program for training village based health auxiliaries and the community health volunteers (CHV) scheme3. Though various attempts were made during last five plans in order to make the health services accessible to rural masses, the actual beneficiaries of health services were found well-to-do urban households. It was felt that, heavy emphasis on hospitals, specialization and super specialization and highly trained doctors are the major reasons for making the health services urban oriented. The sixth five-year plan (1980-85) took these drawbacks as major points and re- emphasized on the community based health care system with special objective of providing health services in rural areas on a priority basis. In order to facilitate this it was planned to train the larger cadre of first level health workers selected from communities. No further expansion of urban and curative health services was emphasized. The seventh plan (1985-90) re-emphasized on the rural health services and laid out its objectives keeping in view the goal of “health for all” by 2000 AD as mentioned in national health policy (NHP), 1983. The plan emphasized on strengthening the rural health services through: the consolidation of the health infrastructure; the strengthening of the three-tier system of sub centers, primary health centers and community health centers; the extension of the MPW scheme with emphasis on training for attitudinal changes; the promotion of community participation; and the involvement of voluntary agencies in the delivery of health services. Despite the rural emphasis, the rural health services for delivery of primary health care were still not operationalized fully. Urban health services, specifically urban slums required immediate attention due to changing urban morphology. Also it 3 The CHVs were part-time workers selected from villages and trained for three months in simple preventive and curative skills both allopathic and indigenous. MPWs supported them, and the program was started in 777 selected PHCs where MPWs were already in place.
- 24. 9 was observed that most of the health care institutions that were established in rural and urban areas in different plan periods were lacking basic facilities such as manpower, drugs and equipment. In addition, lack of proper monitoring of the progress of various programs was found a major hindrance on the way of implementation of different schemes. Accordingly, the eighth plan (1992-97) aimed at correcting these shortcomings through special emphasis on rural and urban health services. For rural areas, the plan aimed at equipping the hospitals with better physical facilities, manpower, materials, equipment and drugs. The urban health services were given due importance with reference to the urban poor and slump dwellers. Under this scheme the urban hospitals were entrusted with the task of running primary health care centers and treat them as their extension counters for providing basic health services to the urban poor communities. Along with the emphasis on rural health services, the plan also aimed at strengthening the secondary and tertiary health care services. 1.2.3. Present health status From the above analysis, it can be inferred that during the post independence period significant effort is being made in India for improving the health status of her population. Therefore, the life expectancy of the population has increased from 44 years in 1960 to 58 years in 1990. The crude birth rate (CBR) as well as the crude death rate has declined from 34 to 32 and 10 to 8 persons during the same period (Table 1.2 and 1.3). It is to be noted that, the current levels of effort and resource allocation are not sufficient to meet the future needs of the people. The higher incidence of chronic diseases like AIDS, hepatitis B and diabetes, and the problems due to new drug resistance for diseases like tuberculosis and malaria adds a new dimension to current health problems. Although provision of these services requires adequate government funding, the degree of government involvement is still relatively limited. Under the present system, India spends 6% of her GDP (US $ 13 per capita per year) on health care annually. While the private households pay 75% of this total health expenditure, the public sector (central and state governments) contribute 22%, which amounts to only 1.3% of GDP (US $ 2.7 per capita). Out of this meager amount, the share of central government is only 10%. A larger share (87%) is paid by state governments, with the third party insurance contributing to the remaining 3% (Krishna Kumar 1993). In addition, the internal allocation of health funding shows that, a major chunk of the government
- 25. 10 expenditure is being spent on salaries. The expenditure on salaries grow at about 10% in real terms while the other health care inputs grow at about 5% rate. Since much of the health care funding is spent on salaries, there is deficiency of funds for non-salary recurrent costs. Expenditure on maintenance of existing infrastructure (including replacement cost) has declined from 30% to less than 20% in the past thirty years. Therefore, infrastructures are not always maintained at an operable level (Duggal et al, 1995). Two policy measures are generally suggested in this regard. First, using the available resources more efficiently; second, exploring the possibility of alternative sources of funding the health services. The present study aims at analyzing these two points in the context of secondary level health care system in Andhra Pradesh. Table 1.2: Health indicators of India, 1961-96 Year CBR CDR IMR LER R U T R U T R U T T 1961-70 NA 41.2 NA NA NA 19.0 NA NA 146.0 * 46.4 1971-80 33.14 28.6 34.7 16.2 9.4 14.8 138.5 79.6 128.8 54.0 1981-89 34.3 27.3 32.7 12.6 7.7 11.5 108.4 62.3 99.6 NA 1990-93 31.0 23.9 29.4 10.6 6.5 9.7 85.0 50.3 78.3 NA 1996 29.3 21.6 27.5 9.6 6.5 9.0 77 46 72 61 * Corresponds to 1951-61 CBR: Crude birth rate, CDR: Crude death rate LER: Life expectancy rate, IMR: Infant mortality rate, R: Rural, U: Urban, T: Total, NA: Not Available Source: SRS Bulletins, reproduced in Health Monitor, 1995. Table 1.3: Life expectancy at birth (years) by sex and residence Year Rural Urban Total Male Female Male Female Male Female 1970-75 48.9 47.1 58.8 59.2 50.5 49.0 1976-80 51.0 50.3 59.6 60.8 52.5 52.1 1980-85 54.0 53.6 61.6 64.1 55.4 55.7 1985-90 56.7 56.2 62.0 64.9 57.7 58.1 1991- 96* NA NA NA NA 60.6 61.7 * Source: India’s Social Sector (1996), CMIE, Bombay. Source: SRS Bulletins, 1970-75, 1976-80, 1981 – 85, 1986 – 90
- 26. 11 1.3. Health status in Andhra Pradesh The health care delivery system in the state of Andhra Pradesh (AP) is little better in comparison to some other states of India. This is true when we look at the health status indicators for the state. In 1996, AP had a birth rate of 22.8, death rate of 8.4, infant mortality rate of 65 per 1000 live births as compared to the corresponding national figures of 27.5, 9.0 and 72 respectively. Life expectancy for AP in 1991-96 was 62.4 as compared to 61.0 for India as a whole. Nearly forty one percent of the births in AP takes place in health institutions or were attended by health personnel which is very close to the corresponding national figure of 41.2 %. The decennial growth rate of 24.2% of AP population during 1981-91 compares favorably with India's figure of 23.8%. At the same time, the total fertility rate of AP in 1994 at 2.5% was slightly less than that of 2.7% for the country as a whole (Table 1.4). Among the common causes of death in AP are nervous and circulatory system disorders (20%), respiratory disorders (16%), maternity and infancy related causes (11%), accidents and injuries (10%), digestive disorders (7%), and various types of fevers (6%). The records from primary health centers (PHCs) show that common illnesses treated were respiratory infections, malaria, alimentary diseases, and aches and pains. Gastroenteritis, respiratory infection, malaria, scabies, and skin infections are among the most frequently treated diseases among children (Mahapatra and Ramana, 1995). Socio-economic variables, which impinge on health development, are given in Table 1.5. The improved health status of AP is closely linked to its well-developed health care delivery system. The public health care system in AP is divided into three levels. Primary care facilities, which provide the people with preventive and promotive care for minor health problems, maternal and child health (MCH), and family planning (FP). The Directorate of Health Services (DHS) manages this section of health services, with the exception of family planning, which is under the control of Directorate of Family Welfare (FW). The referral hospitals and secondary level hospitals make up the second level of public health care. These facilities provide both inpatient and outpatient care for illness that are too complicated to be treated at the primary level. These facilities are under the control of the Andhra Pradesh Vaidya Vidhana Parishad (APVVP), an autonomous governmental body, which was created in 1986. Tertiary hospitals, which include teaching hospitals, are the third and final level of public health
- 27. 12 system. Managed by the Directorate of Medical Education (DME), the tertiary hospitals provide more technical and specialized care. Table 1.4: Health indicators of Andhra Pradesh and India Health indicator Location AP India Annual birth rate per 1000 population , 1996 R 23.5 29.3 U 20.6 21.6 T 22.8 27.5 Annual death rate per 1000 population, 1996 R 9.2 9.6 U 5.9 6.5 T 8.4 9.0 Infant mortality rate per 1000 live births, 1996 R 73 77 U 38 46 T 65 72 Estimated death rate for children aged (0–4) years, 1993. R 19 26.6 U 11.9 13.4 T 17.1 23.7 Life expectancy, 1991-96 T 62.4 61.0 Attended births (%) by qualified personnel / institution (per lakh population), 1993 R 56.2 43.5 U 90.6 88.1 T 62.9 48.8 Decennial growth rate, 1981-91 R 18.4 20.0 U 43.2 36.5 T 24.2 23.8 Total fertility rate, 1994. R NA NA U NA NA T 2.5 2.7 R: Rural, U: Urban, T: Total Source: 1. SRS Bulletins, 1988, 1994, 1998 2. Final population totals: Brief analysis of primary census abstract (1991, 1992), Census of India, Series –1, Paper –2. 3. Heath Information of India (1992), New Delhi. 4. India’s Social Sector (1996), CMIE, Bombay. Apart from the directorates that manage the individual levels of the public health system, there are five smaller directorates that manage the AIDS program,
- 28. 13 Preventive Medicine, Employees State Insurance, and the Indian System of Medicine and Drug Control. The Department of Health, Medical, and Family Welfare is responsible for these smaller directorates. Table 1.5: Socio economic indicators of Andhra Pradesh and India Indicator Reference period AP India Real per capita income (Rs.) 1990-93 1829 2219 Relative index of development 1992-93 99 100 Index of infrastructural development 1992-93 103 100 Total non-agricultural workers in labor force (%) 1981 30.5 NA 1991 31.4 NA Female non-agricultural workers in labor force (%) 1981 17.1 NA 1991 17.3 NA Urban population (%) 1991 26.9 25.7 Rate of literacy (%) 1991 44.1 52.1 Rate of female literacy (%) 1991 32.7 39.3 Drop-out rate from classes I – IV (%) 1989-90 56.2 48.8 Average size of operational holdings (ha) 1985-86 1.72 NA Road length per 1000 sq. km. 1987-88 500 544 Electrified villages (%) 1991 100 84 Source: Basic Statistics: States (1994), CMIE, Bombay. Profile of Districts (1993), CMIE, Bombay. Final population totals: Census of India 1991, Series-1, Paper-2. As of 1994, AP's public health infrastructure includes 10,555 sub-centers (SCs), 1,306 primary health centers (PHCs), 175 community health centers (CHCs), and 45 mobile units at the primary level. At the secondary level, there are 162 hospitals and dispensaries of which 107 are community hospitals, 8 are area hospitals, 17 district hospitals, 6 specialty hospitals and 24 dispensaries. While the community hospitals have 30 or 50 beds and are in rural areas, area hospitals have 100 beds and are found in smaller towns. District hospitals located in district headquarters have nearly 200 beds or more. All dispensaries are situated in urban areas. For tertiary care, 38 tertiary hospitals affiliated with 9 medical schools, and 2 super-specialized hospitals are available. In addition to these facilities, there are 23 district TB centers, 25 TB clinics, and 194 leprosy
- 29. 14 control units at the primary and secondary levels. Apart from use of government health care facilities, utilization of private health care facilities happens to be higher for the state of AP. According to the national sample survey (NSS) utilization data from the 42nd round (1986-87), there seems to be a relatively higher utilization of private hospitals and private doctors as far as outpatient (OP) care was concerned. Private health care caters to a higher extent of OP cases (74.9%) in the rural areas of AP than in the urban areas (71.9 %). In OP care at all India level, private health care institutions are the main providers of care, attending to 69.2% of cases in rural areas and 69.9% cases in urban areas. In AP, 69.6% of inpatient (IP) cases in rural areas and 61.6% in urban areas are treated in the private health care institutions. Public health care utilization for IP care in rural AP is only around 30% and in urban AP around 38%. The very high utilization of health care in the private sector in AP is in contrast to the national picture. In India, in the rural areas 59.7% of IP cases were treated in public hospitals or PHCs and an almost similar percentage (60%) in urban areas. Lack of access to public health care facilities, low quality and non-availability of staff, drugs, and essential supplies make public facilities an unattractive choice for general population. Scheduled Tribe (ST) households and the lower-income groups are the major users of public health facilities. Since 98% of the public health services are free, as opposed to only 7% in the private sector, it is rational for the poorer people to use more government facilities. Two- thirds of the people who prefer public health care to private care are in the lowest 25 % of the income bracket (Mahapatra and Ramana, 1995). 1.4. Allocation of funds to health sector in AP In the past couple of years, the amount allocated to health care in AP has declined. Public health care spending in AP constitutes only 5.6% of the total state revenue and 5.1% of the total state revenue and capital budget, which represents 1.3% of the State Domestic Product. The pattern of allocation of funds within the health care system has also changed over the past 10 years. While expenditure allocated to primary health care has increased from 46% to 49% between financial year 1981 and 1990, funding to hospital services has declined from 41% to 34%. In the amount allocated to hospital care, most of it goes to tertiary care, leaving very little resources to the secondary care despite the planning commission norm of 33:67. The ratio is as high as 49:51 between tertiary and secondary care respectively. With only 32% of its public hospital
- 30. 15 beds at the secondary level, AP public health system does not meet even half of the Planning Commission norm of having 70% of all public hospital beds at secondary level. Moreover, shortage of diagnostic facilities, repair and maintenance service for equipment, transport facilities, and trained personnel are very often a feature of secondary level health care. Inadequate beds and technical equipment has hindered the secondary level's capacity to support primary level health care when needed. Overall, shortage of hospital beds can also be observed in AP. According to the Planning Commission norm, 67 beds should be available for every 100,000 people in the population. Based on this norm, there is a shortage of approximately 12,000 beds. However, this is not always the best indicator of the needs of the society. A better way to assess the need for hospital beds is by analyzing the epidemiological profile. Based on this analysis, 90,000 beds are needed to maintain a minimal level of clinical service in AP. Even after the private sector provides 42,000 beds, there will still be a shortage of 17,000 beds (Mahapatra and Ramana 1995) The above paragraphs suggest that, in spite of better health care facilities in AP, the usual problem of low quality, lack of access, non-availability of required staff and essential drugs still persist. Since these problems stem mainly from lack of efficiency and inadequate availability of public finance, there is a need for looking at these two aspects in detail. 1.5. Motivation for the present study The secondary level health care system in the state of AP gives an unique opportunity for a detailed study of these two aspects. The uniqueness lies in the fact that, a separate governmental autonomous body manages the secondary level hospitals in the state of AP, called Andhra Pradesh Vaidya Vidhana Parishad (APVVP). The autonomy status to the secondary level hospitals is a first of this kind in India. The aim of granting this autonomy status was to provide the health services with greater efficiency, quality and improvement in financial sustainability, thereby minimizing the inherent problems that are associated with other health care delivery system. The major policy question here is, how far these objectives have been useful in improving the delivery of secondary level health services in AP. Are the secondary level hospitals running efficiently? Are the hospitals self sufficient in funds in order to maintain better quality of care?
- 31. 16 The present study aims at examining these issues from an economist’s point of view. Explicitly, the present study has two broad objectives: 1.6. Broad objectives of the study: (a) To evaluate the performance and efficiency of secondary level hospitals in Andhra Pradesh. (b) To examine the issue of health financing in developing countries in general and AP in particular, with a view to suggest some policy measures in this regard. 1.7. Chapter outline: As a background of this study, in this Chapter, we made an attempt to give a brief outline of the health care delivery system of India as well as Andhra Pradesh. Based on the problems identified with the health-care delivery system in Andhra Pradesh, the broad objectives of the study are also specified here. In Chapter II, a selected review of the studies on evaluating hospital performance and efficiency that are conducted in developed and developing countries is presented. Chapter III deals with the methodology of past studies on the subject. We also try to identify the appropriate methodology for this study. Based on the review and methodology, detailed objectives of our study are listed. In Chapter IV, an attempt is made to evaluate the performance of secondary level hospitals in Andhra Pradesh through the use of combined utilization and productivity (CUP) analysis. Chapter V is devoted to evaluate the efficiency of secondary level district hospitals in Andhra Pradesh using cost functions approach. Chapter VI deals with financing aspects of health services. Here, we give a brief outline of different methods of financing the health services in underdeveloped countries with a focus on the government financing, its inadequacy, and the need for additional resources. Further, the possibility of charging the user fees for curative services is tested based on a field survey in one of the AP secondary level district hospitals. Chapter VII gives the summary of the main findings of this study, its drawbacks and the scope for further research.
- 32. 17 CHAPTER – II REVIEW OF LITERATURE 2.0. Introduction This chapter gives a brief over view of selected studies on efficiency of health services and health sector financing with a view to understand the issues involved and identify the appropriate methodology for this study. The review is made non-technical, keeping the technical details to the next chapter. The studies relating to health sector efficiency are reviewed separately for developed and developing countries. This is followed by studies on Andhra Pradesh. In the penultimate section, we give a brief review of the health financing issues and studies that stress on user charges. The last section contains conclusions of the chapter. 2.1. Studies on hospital efficiency 2.1.1. Studies using performance indicators Lasso (1986) analyzed the performance of 73 Colombian hospitals for the period 1977-80. Hospitals are categorized into different groups on the basis of their bed strength. Using combined utilization and productivity (CUP) analysis for assessing their performance, the author found that hospitals with larger bed strength have higher BOR and ALS followed by other groups with smaller bed size. Sear (1991) conducted a study of efficiency of investor owned and non-profit hospitals by using a sample of 142 hospitals in the State of Florida, USA in1988. Three indicators (ratios), namely total number of full time equivalent (FTE) personnel per active bed, number of man hours per adjusted patient day, and the total wages paid per adjusted patient day are used for the purpose. The author found that investor owned hospitals used significantly fewer FTE staff per bed, fewer man hours per adjusted patient day and paid significantly less in wages.
- 33. 18 2.1.2. Econometric studies There are a number of studies, which used econometric models for studying hospital efficiency. The problems addressed here include, quality and case mix heterogeneity, uncertainty, and input price variations. The nature of these problems and the way they can be tackled are briefly discussed below in the context of developed country. The review of studies for both developed and developing countries follows thereafter. (i) Quality heterogeneity Most studies on health services ignore the problem of quality heterogeneity. Quality variations in health services hinder efficiency comparisons between providers. Failure to control quality differences may ascribe higher efficiency to lower quality producers and vice-versa. The next question is what is health care quality and how it can be measured and controlled for? There are alternative definitions for health care quality. Wyszewianski et al (1987), for example, adopted a conventional, technical definition of health care quality. It is based on both the extents to which health care providers comply with technical norms of care and the service’s ability to improve the patient’s health status. According to them, “ … quality is higher for care A than for care B, if for the same patient, care A is likely to make a greater net contribution to the patient’s health and wellbeing than is care B.” Barnum and Kutzin (1993) recognize that the quality of health care has two dimensions, the medical or technical and the consumer’s dimension. “Quality has both supply and demand characterastics. The critical demand issue is ‘perceived’ quality: the consumer assessment of the relative quality of different health care providers…. Adequate staff and supplies are obvious supply side factors affecting ‘actual’ quality of services that are important in affecting perceived quality” (Chapter 3, p.17). The notion of quality that we adopt for our study is demand side one. This includes the demand side dimensions such as attitude of medical personnel, cleanliness of premises, and other measures of the degree to which patients’ preferences are met. Because of the multi-dimensional nature of quality, it is not possible to infer the estimates of efficiency by combining quality information with
- 34. 19 cost of care. To do so, knowledge about the weights of different dimensions of quality would be required. That, however, is beyond the scope of this study. It is sometimes argued that perceived and technical quality may or may not be in agreement with one another. Discrepancy in these viewpoints can be greater in developing countries where poor, uneducated populations may have a more limited ability to discern the technical merits of health services. Barnum and Kutzin (1993) suggest that quality perceptions by patients are in some cases related with technical quality measures, such as the availability of drugs and medical supplies. Thus, the policy measures aimed at improving technical quality through increase in supplies can also improve quality perceptions. To sum up, the quality of care can be assessed and controlled for in various ways when measuring efficiency of health services. These include measuring quality according to (1) degree of compliance of medical norms of practice, (2) clinical outcomes and (3) various measures of patient’s satisfaction. (ii) Case mix Comparative studies of health facility cost and efficiency have to overcome the difficult methodological problems arising from the wide diversity of health care activities produced by alternative providers and the effects that such heterogeneity has on resource use, cost, and efficiency. A patient-day care may differ between hospitals, between different departments of the same hospital and over time. According to Tatchell (1983), these differences may arise due to differences in (a) technology, (b) quality of care, (c) case-mix, (d) case complexity and severity, and (e) institutional characteristics (size, teaching status, location, composition, ownership etc.). So the output of the hospitals need to be standardized in order to bring it to a comparable form for measuring efficiency. Among the different methods of standardising the output, service mix is the foremost. Service mix may be defined as the type of services available or the services and the procedures actually performed. Cohen (1967), Anderson (1976), Kouner (1969) have used the method of service-mix, taking the actual procedures performed in the hospital. Another method of standardizing hospital output is to derive a measure of hospital case-mix using the case types treated in the hospital. One of the early
- 35. 20 attempts to standardize hospital output for case mix was based on specialty mix. Feldstein (1967) made an attempt to standardize the hospital output by dividing the various specialties into 8 mutually exclusive groups. Another approach to deal with case mix is diagnostic mix. The studies related to this latter method have used either international classification of disease (ICD) groupings or diagnostic related groupings (DRG) for standardizing the output. Evans (1971) and Evans and Walker (1972) have used ICD groupings for dividing the diagnosis into different homogenous categories. Fetter et al (1980) and Barer (1982) have used DRG for standardising hospital output. Other methods of standardising hospital output are found in Lave and Lave (1970), Lave, Lave and Silverman (1972). However, Tatchell (1983) gave a detailed review of the studies related to the issue of case mix and service mix. He identifies two methods of standardizing hospital output, namely service mix and case mix. The service mix approach uses information about the type of services offered by the hospital (pediatrics, gynecology, etc.) as a basis for standardising hospital output. Tatchell (1983) argues that the existence of common services among providers does not imply that the mix and complexity of cases, as well as quality of care are similar among providers. He also points out that whether or not certain services are offered by a provider is not sufficient to characterise its output, due to lack of information about the extent to which these services are actually used. The case mix approach attempts to standardise hospital output according to the mix of cases actually treated in the hospital. Vitaliano (1987) calls this method as demand-side one. In fact, both methods reflect demand and supply forces since the services offered by hospital generally respond to a demand. The case mix approach, however, captures supply as well demand relations better because it is based on utilisation information. Under the case mix approach, Tatchell distinguishes several methods. They are; specialty mix, diagnostic related groupings, information theory, case severity, and others. The reminder of this section focuses on the use of the case mix approach to standardize output when measuring health facility efficiency. The case mix denotes composition as well as complexity of cases. While measuring hospital cost, these two aspects are of utmost importance and need to be considered. For example, if only composition is taken into account, then wards in two hospitals would appear identical if both of them treat same number
- 36. 21 of diabetic cases over a given period of time. If information on case complexity is also taken into consideration, it could be that the two facilities differ in terms of the proportion of complicated cases. Also it is to be noted that the facilities, which treat more number of complex cases consume more resources resulting in higher cost per treatment. Thus, study of health facility efficiency, technical or economic, requires knowledge about the case mix. This information can be used to control for case mix variability while estimating cost. However, while it is possible to obtain data about composition of output, it is usually difficult to get information on output complexity. Complementary techniques must then be used. Some of these are given below. Register and Bruning (1987) in their comparative study of efficiency between profit and non-profit oriented U.S. hospitals have used an alternative method for dealing with case mix, since the information on case mix of individual hospitals was not available. Case mix proxies are used to limit the confounding effects of case heterogeneity on efficiency measures. They assumed that rural and urban hospitals differ in their case mix and therefore confined to only urban hospitals in their study. All long-term (older) federal hospitals are also eliminated from the sample to reduce heterogeneity. Hospitals that provide a specified set of services are retained, thus eliminating all hospitals, which are too `basic’ or `high tech’. It is posited that hospital size, measured by the number of beds, is also associated with the case mix. Therefore, they limited their sample to hospitals within the range of 100 to 200 beds. To verify their hypothesis that hospital size is associated with case-mix, the authors group the sample hospitals by size and compare the groups on the basis of bed-to-doctor and bed-to-nurse ratios. The differences in ratios are interpreted as differences in case mix. However, it should be noted that the above procedure might introduce a bias in their analysis because bed-to-labor ratios may reflect not only case mix but also efficiency differences. The authors have also estimated technical efficiency using multiple-output production function approach. This estimation method allowed them to control case mix composition, while the previous procedure of restricting the sample could control the case complexity as well. Eakin and Knieser (1988) in their study of economic efficiency of US hospitals also confronted the problem of case mix differences among facilities. To quote
- 37. 22 them: “The output data indicate that hospitals are a heterogeneous group with respect to mix of cases treated. Consequently, we adopt a multiple output specification in this study. In particular we use four categories of outputs - general medicine, obstetrics, gynecology, weighted surgeries and outpatient visits” (p.587). The above procedure controls for case mix composition. In order to control for case severity and factors other than case mix the authors proceeded as below: “Three additional variables – TEACH, a dummy variable indicating medical school affiliation, ALOH, the average over all length of hospitalisation and the average overall occupancy rate – are included in the list of repressors to control for case severity and capacity utilisation” (p.587). The variable TEACH is used to capture the (positive) effect on hospital costs of teaching activities. The length of stay variable is used to control for severity. However, this variable may represent differences in medical practice, not attributable to severity but to efficiency. The occupancy rate variable is introduced to discriminate between technical and economic efficiency, a useful distinction. A hospital may be technically efficient (it can be on its production possibility frontier) but, due to low occupancy it may be economically inefficient. Vitaliano (1987) adopted a service mix approach to control for case mix and its effect on hospital efficiency. For each hospital in the sample, an un-weighted index of 11 available hospital facilities or services is constructed and used as an independent variable to estimate a total cost function. He finds that the service- mix variable is statistically significant and has a positive effect on total cost. To sum up, there are several methods available to control case mix when evaluating health services efficiency. These include: (a) limiting the sample of providers to those that provide a similar case mix, (b) in econometric studies of efficiency, include indicators of case mix such as type and volume of services produced, measures of output complexity etc. as explanatory variables and (c) limiting the comparison of efficiency to a few, well defined medical services (e.g. appendectomies, hernia repair, etc.).
- 38. 23 The author has also mentioned several other methods under case mix approach, such as specialty mix, DRGs, information theory, case severity and others. (iii) Uncertainty The uncertainty in the hospital care arises from different dimensions. These are uncertainty with respect to (a) the timing of an illness and hence the demand for hospital services, (b) physician's diagnosis, and (c) the final out-come of the treatment. Since demand uncertainty in particular is likely to have important consequences on hospital costs, it is necessary to understand the cost and uncertain demand. If one views the hospital, in addition to providing direct or anticipated patient care, as also providing sufficient capacity to assure that hospital services are available at the time of unexpected demand, this insurance or option demand of the community at large should be treated as yet another service provided by the hospital. Failure to take this standby service into account in econometric cost models may result in specification error. Cowing et al (1983) suggest the following formula for measuring the level of standby services: 1/2 Census)Daily(Average Census)DailyAverage-(Beds k … (2.1) Here, larger the ‘k’, there is less chance that services will not be available for yet another patient (Jaskow, 1980). Although this issue has been recognised by various authors, it could not be incorporated in hospital costs in a more formal manner. A detailed review of the studies dealing with this issue is given in Cowing et al (1983). In brief, it suggests that demand or output uncertainty accounts for some portion of the observed excess capacity of hospitals and it had lead to decreasing cost with increase in hospital utilisation. (iv) Input price variation Input price variation across providers complicates the interpretation of differentials in efficiency. For example, let there be two providers who operate with equal technical efficiency, but facing different input prices. An analysis of
- 39. 24 efficiency that focussed on technical efficiency alone would conclude that both the providers are equally efficient. This would contrast with the outcome of a study on economic efficiency for the following reasons. If the technology and input mix were similar between providers, the provider facing higher input prices would exhibit the lower economic efficiency. Furthermore, if factor prices vary among providers, no inferences can be drawn about economic efficiency, unless a more careful analysis is conducted. Social cost of resources, instead of factor prices, may have to be considered in such cases. There are numerous studies which examined some of the above mentioned issues in detail. These include, Bays (1980), Pauly (1978), Baron (1978), Elnicki (1974), Lee Wallace (1973), Lave, Lave and Silverman (1972), Hu (1971), Francisco (1970), Cohen (1967), Cowing and Holtman (1983), Watts and Klastonin (1980), Luke (1979), Feldstein and Shutting (1977), Bays (1977), Thompson, Fetter and Mross (1975), Rafferty (1971), and Berry (1970). 2.1.3. Studies for developed countries During the past two and half decades, estimation of cost function for hospitals has become a common place for industrialised countries. Reviews by Cowing, Holtman and Powers (1983), Wagstaff (1989), and Ricardo (1992) have documented this literature. The initial attempts to estimate cost functions using data from hospitals in industrialised countries employed composite measures of hospital output (e.g. Cohen 1967). They used average or unit cost of inpatient day or admission as the dependent variable, and a variety of interrelated explanatory variables such as occupancy rates, patient flow, length of stay, and capacity as explanatory variables (Mann and Yett 1968). Most of the early empirical studies [Feldstein (1967), Berry (1967), Carr and Feldstein (1967), Ingbar and Tayler (1968), Fransisco (1970)] were concerned with the effect of size on hospital costs – that is, the existence of scale effects and the related questions of optimal size. Authors of these studies estimated rather restricted versions of the general cost function, usually without recognizing many of the implicit assumptions involved. For example, most of the models did not include input prices in the estimated cost function, an omission which is equivalent to assuming zero input substitutability, that is, a fixed proportions technology for producing hospital care.
- 40. 25 In addition, many of these studies typically include two types of variables in the cost function. The first one is a flow type variable, number of patient days of service provided, to account for variations across hospitals in actual or current output. The second variable is a stock type variable, the total number of beds, to account for variations in size or capacity. Some models have used other measures of output and the ratio of output to capacity. The basic purpose of these latter models appeared to be that of relating cost changes caused by output changes, holding capacity constant, to the short-run cost structure; and cost changes caused by changes in capacity, holding output constant to the long- run cost structure. This in turn is related to the existence of economies of scale. The general conclusion of these studies was that there was evidence of significant economies of scale, at least up to moderate sized hospitals of around 500 beds. Later studies have specified cost functional forms and included variables that are consistent with a theoretical production structure (e.g. Cowing and Holtman 1983, Grannemann et al 1986, and Vita 1990). They have used total cost (rather than average cost) functions with multiple outputs and have employed flexible functional forms. The data used were mostly from USA or UK. We discuss briefly some of these studies. Granemann et al (1986) departs from previous studies of hospital cost by introducing a more flexible functional form for total cost. They combined a translog form with a conventional cubic cost function where outputs enter directly, not in the logarithmic form; to permit zero output levels. Capital is included as an independent variable, reflecting their assumption that capital is not a fully fixed cost to the hospital. This allows exogenous variations in the hospital’s capital cost even in the short-run. They also included various hospital outputs, like emergency care, home visits, and two measures of inpatient care, to account for differences among hospitals in average length of stay. Due to data limitations, they have not included the interaction terms between input prices and output levels in their model. The authors estimated various measures of hospital performance such as marginal cost, product specific economies of scale and scope. Data on hospital costs, inputs and outputs are taken from American Hospital Association’s annual survey for 1981. Ordinary least squares estimation method was used for estimating the cost function. Vitaliano (1987) argued that econometric studies of efficiency, which use multiple
- 41. 26 outputs in a cost function are plagued by the problem of multicollinearity among outputs. He, therefore, used a total cost function with a single output to conduct a comparative study of economic efficiency using a sample of 166 New York State hospitals. Instead of using output measure as an explanatory variable, the author used the number of beds. Other explanatory variables used in the model include, condition of hospital, index of case mix of services provided in the hospital, regional dummy, and a measure of hospital’s monopoly and monopsony power. The latter variable is measured by hospitals’ share of the total supply of beds in the country. Vitaliano argued that monopsony or oligopsony hospital might exert downward pressure on factor prices. The estimation technique was weighted least squares. A quadratic U-shaped average cost function was obtained from the estimated total cost function. As expected from a U-shaped average cost curve, the hospital costs have exhibited economies of scale in certain range of output. He attributed this to presence of high fixed cost such as specialised personnel and equipment. Eakin and Kniesner (1988) proposed a general methodology for estimating hospital cost functions, which allows for the possibility that hospitals may not be cost minimisers. They point out that “erroneously assuming cost minimisation leads to inaccurate estimates of factor substitution possibilities” (p. 584). The authors estimated a system of equations which include observed cost function and three observed factor share equations. They specify a ‘hybrid’ translog total cost function to accommodate zero output levels instead of their logarithm. The model included four outputs and four inputs and was estimated using the method of seemingly unrelated non-linear regressions. The sample consisted of 331 U.S. short-term hospitals, both profit and non-profit oriented. Allen elasticities of factor substitution and economies of scale are estimated. The empirical results show that hospitals undervalue the cost of capital and over value the price of physicians. Hospitals thus tend to over employ capital and under employ physicians. This resulted in allocative inefficiency, which equals to 5 percent of the total observed cost. They also found that elasticities of substitution and factor demands are sensitive to model specification. Frank and Taube (1987) studied technical and allocative efficiency in the production of outpatient mental health clinics using data from 755 clinics obtained through a survey by the U.S. National Institute of Mental Health in 1982. The study focussed on estimation of production functions for mental health visits. Like
- 42. 27 Eakin and Kniesner (1988), this study was also interested in exploring provider departures from cost minimisation behavior. The method consists of verifying marginal productivity theory, i.e., equality between the factor price ratio and the ratio of marginal products for each pair of inputs. Differences between the above two ratios are attributed to either over or under utilization of inputs. The authors also estimated the economies of scale and scope in the production of mental health visits. Finally, the study looked into the determinants of efficiency by focussing on differences in productivity between government-run and private clinics. In the empirical analysis, Cobb-Douglas and Transcendental functional forms are used for specifying a production function. Ordinary least squares estimation was used. Both the forms indicated decreasing returns to scale and greater productivity of private clinics, as evidenced by an ownership dummy included in the production functions. The study also has shown violation of marginal productivity theory, i.e., factor price ratio between physicians and other clinical staff differed from the corresponding ratio of marginal products, with physicians being over employed. This also signaled a departure from cost minimization in the production of outpatient mental health services in the USA. Eakin (1991) used the results from Eakin and Kniesner (1988) to study the determinants of economic efficiency in hospitals. For this purpose, the estimated values of allocative inefficiency are regressed on several hospital and market- related characteristics considered to be the determinants of efficiency based on economic theory. The independent variables included ownership, regulatory factors, competitive factors, factors characterizing the sources of hospital revenue, measures of hospital size, and regional dummy variables. Estimation was done using generalised least squares. It was found that the regulatory environment is a critical determinant of economic efficiency. Economic inefficiency was greater in larger hospitals with a larger market share. He does not find statistically significant differences in hospital efficiency between non- profit and profit-oriented hospitals and like-wise between church operated and other hospitals. 2.1.4. Studies for developing countries Initial studies on hospital cost and performance in developing countries have made use of accounting information and hospital service records. A detailed
- 43. 28 review of the Indian studies on cost accounting method prior to 1974 can be found in Srivastava and Gupta (1974). Sharma and Timmappaya (1974) attempted to develop a suitable method of measuring the performance of smaller hospitals. Ramlingaswami (1984) estimated the cost of medical education in 17 medical colleges in India and found that the recurrent cost of educating a medical graduate varied between Rs. 55,000 to 1,34,000 in the year 1981 among different medical colleges. Thus, most of the earlier studies have concentrated in developing a suitable methodology for costing rather than evaluating hospital performance. Rodriguez and Jimenez (1985) conducted a comparative study of productivity between private and (decentralized) public Chilean hospitals. They measure the productivity in terms of inverse average length of stay (ALS), namely shorter the ALS, higher the productivity. The authors divided ALS into three components, namely diagnosis (D), medical treatment (T) and recovery (R) respectively. It was argued that while T must be performed within the hospital, D and R can partially be accomplished on outpatient basis. For a given patient-case-mix, the authors hypothesized that a series of individual specific variables, such as age and income can influence D and R. It was assumed that a patient’s income is a close proxy of the type of health insurance or coverage he / she has. Thus, income was assumed to closely correlated with out-of-pocket price of the services and it influenced individual behavior, as measured by D and R. The type and severity of illness, and the amount of medical inputs provided to the patients, were also assumed to affect the D, T and R. Using a sample of 5 hospitals (3 private and 2 public) and 369 patients, the authors used OLS to regress length of stay on patient age and income, medical inputs and facility specific dummy variables. Three separate equations were estimated for obstetrics and gynecology, surgery and internal medicine. The authors conclude that, other things being equal, private hospitals had lower length of stay and were more productive than public institutions. They cautioned, however, that their analysis was subjective to the accuracy of information about diagnosis, case mix and severity of the illness etc. Lewis, Suletta and Forgia (1990) used cost accounting method to measure efficiency and quality of care at Aybara hospital, which is a 271-beded
- 44. 29 government-run facility in the Dominican Republic. To estimate costs, the authors monitored a selected sample of patients during their treatment in the hospital, recording the cost of services provided to them. The sample consisted of three sets of patients, (i) emergency patients, (ii) people consulting in an ambulatory basis, and (iii) inpatients admitted to five departments (3 surgical and 2 ophthalmology) of the hospital during a one-week reference period. The sample patients underwent 23 surgical interventions consisting of wounds, appendicitis, cataract, hysterectomy and hernia during that week. The prices of non-labor inputs were obtained directly from supplier wherever possible. Labor costs were computed by measuring actual staff time devoted to medical procedures and multiplying it by workers actual wage (converted to an hourly basis). The total cost of procedures was the sum of the variable and allocated fixed cost. The variable cost included cost of labor, drugs, ancillary services and consumables. Cost of overheads, depreciation of buildings, equipment and other fixed assets constituted the fixed cost. The quality of care was measured in two ways, (a) assessing the appropriateness of the qualifications of the medical staff involved in care and (b) comparing the actual diagnosis and treatment practices and services delivered with medical norms of care. By extrapolating the percentage of labor used from the sample to all patient services in the hospital, the authors concluded that only 12% of the medical labor contracted by the hospital could actually be counted for. Although no allowances were made for down time by the medical staff, this unexpected result signaled a major inefficiency in operations of Aybara hospital. With regard to quality of care, major departures were found between the expected cost of meeting norms of diagnosis and treatment and the cost of such services actually provided. The cost of drugs dispensed and the tests performed represented about 10% of the costs implied by the norms. Anderson (1980) studied the behavior of hospital costs in Kenya using a sample of 51 hospitals during 1975-76. The dependent variable was average cost per patient day. The explanatory variables were, capacity as measured alternatively by available and used beds, occupancy rate, average length of stay, number of outpatient visits per inpatient day, number of satellite ambulatory facilities operating under the hospital and the nature of hospital, provincial or non- provincial. Four alternative specifications of the model were estimated using alternative
- 45. 30 dependent and independent variables. It was found that the hospitals were operating with increasing returns to scale, as evidenced by negative and statistically significant coefficients associated with scale variables. Further, higher occupancy levels resulted in lower average costs implying that greater demand should be accommodated within the existing hospitals rather than through new ones. Outpatient activity was found to increase average cost. In contrast, length of stay did not come out statistically significant. Because the cost of hospital and its satellites are intertwined, the average cost used as dependent variable also included the cost of satellite facilities. The regression results showed that a greater level of satellite activity had a positive impact on aggregate average cost. Finally, the provincial hospitals were found to have higher average cost than district and sub-district hospitals. Dor (1987) has estimated average cost function for 19 urban hospitals in Peru. The hospitals included were some from the ministry of health and others from social security ministry/department. Using the average cost function, Dor obtained an analytical expression for the optimal patient flow, i.e., the flow at which hospital average cost was a minimum. Using OLS and WLS methods, he estimated separate average cost functions for each of the three cost categories (labor, goods and services) as well as combined. He found that hospital average cost decreased with service intensity. This means, the average cost of hospitalization decreased either with increase in number of hospitalizations or decrease in the number of beds or both. The calculated optimal flow was 3.2 admissions per bed per month. This value was above the sample mean but below the actual flow of several hospitals in the sample. Thus, the study concluded that the average cost curve was ‘U’ shaped. Bitran and Dunlop (1990) studied the determinants of hospital costs using a sample of 15 government hospitals for Ethiopia, with one to three annual observations for each hospital and total of 38 observations. Flexible functional form as in Grannemann et al (1986) was estimated using the OLS method. Several parameters of interest such as the marginal cost of inpatient care and outpatient care, average incremental cost, product specific economies of scale, and economies of scope measures were calculated from the estimated cost function. The authors found that the hospitals in the sample were operating under constant returns to scale for patient days, laboratory tests, and deliveries. Economies of
- 46. 31 scope were found between first outpatient visit and inpatient days, signaling an economic advantage (reduction in cost) in joint production of inpatient and outpatient care. The limitation of the study was to use ‘planned’ hospital expenditure by the Ethiopian government to represent cost. This may differ substantially from the actual expenditure. Wouters (1990) studied the cost and efficiency of a sample of 42 private and public health facilities in Ogun State, Nigeria. She analysed efficiency and cost by estimating production and cost functions respectively. Technical efficiency was assessed using the estimated production function and the associated measure of marginal product of health workers. She found that the efficiency variable was insignificant and thus concluded that departures from cost minimisation have little effect on expenditures. She also found that the marginal costs are less than average costs and thus concluded that facilities in her sample exhibit increasing returns to scale both for admissions and outpatient visits. Based on the estimated economies of scope, it was concluded that there were no advantages in the joint production of inpatient and outpatient care. Shephard et al (1991) conducted a cost effectiveness study of surgery in intermediate health units (IHU) in Cali, Columbia. In order to reduce the high cost of treatment, IHUs were set up as an intermediate unit between PHCs and hospitals. The study showed that surgical anesthesia complications were higher in the hospital than in IHU, patient’s satisfaction as measured by the time taken to return to their work after surgery, was higher at IHU. The average cost was much higher at the hospital than in IHU. Barnum and Kutzin (1993) computed the same indicators as in Lasso (1986) for measuring the performance of public hospitals in some of the developing and industrialized countries. The data used by them was either average of all hospitals in the selected countries or average of sample hospitals in respective countries. They found that the industrialized countries perform better in terms of their utilisation as well as productivity compared to developing countries like Turkey, Ethiopia and Korea. Bernum and Kutzin (1993) have also conducted an evaluation study of financial management of 8 Class I and Class II hospitals distributed throughout Columbia. The costs were evaluated at 1975 constant pesos, and service data correspond to each year during 1975-78, thus providing a pooled time series of cross
- 47. 32 sections. Using a total cost function, the authors found that at the sample average, the marginal costs for inpatients and outpatients were approximately equal to the respective average costs, implying short-run constant returns to scale with respect to both the variable factors. Another study was conducted in 1993 by the same authors for China. A short-run variable cost function was used for the purpose. The sample included cost and service data for three years, 1984-86. The authors found diseconomies of scale and only mild economies of scope, implying thereby short-run inefficiency in the level of operation with respect to bed days and outpatient visits. 2.1.5. Studies for Andhra Pradesh: Mahapatra and Berman (1991) used the method of ratio analysis for evaluating the performance of secondary level hospitals in Andhra Pradesh, India. Due to heterogeneity in the hospitals covered, the authors grouped the hospitals on the basis of available number of beds and the type of service provided. The hospitals providing similar type of service are grouped into one category. The authors found differences in the performance of different categories of hospitals. Mahapatra and Berman (1994) applied the combined utilization and productivity (CUP) analysis for evaluating the performance of secondary level hospitals in the state of Andhra Pradesh (AP). Based on a sample of 108-109 hospitals for the years 1989-90, the authors computed bed occupancy rate, bed turnover rate and average length of stay for each hospital in the sample. Using the graphical method suggested by Lasso (1986), the authors found that a sizable proportion of hospitals, namely, 40.7 percent in 1989 and 39.4 percent in 1990, were associated with low productivity and low utilisation. Similarly, an estimated 31.5 percent of hospitals in 1989 and 33.9 percent in 1990 were associated with high utilisation and high productivity. After the study by Mahapatra and Berman (1994), there were no published works, to our knowledge, that measured hospital performance in AP. Since there has been a substantial change in policies relating to secondary level hospitals, there is a need to re-examine the performance indicators of the secondary level hospitals in AP. This is being attempted in this study. Thus, although there are a large number of studies relating to hospital cost in developed countries using the cost function approach, there are only a very few studies for developing
- 48. 33 countries. To the best of our knowledge, there are no published studies in India using cost function approach for measuring the hospital performance and efficiency. The present study aims at filling this gap by using data on secondary level district hospitals in AP. 2.2. Financing of health services Government intervention in the provision of health care services is supported on the grounds of externality, equity and efficiency. Several economists like Arrow (1963), Newhouse (1970), and Cuyler (1971) gave sufficient justification for the government provision of health services. They argue that private providers can not supply the health services due to market failure. Since health services are basic and they help in improving the quality of human capital and thereby economic development, there is a need for the government provision of these services. World development report (1993) reiterates this view. Further, the report states that the government provision of health services helps in reducing poverty, inequity and inefficiency. As already discussed in Chapter I, World Health Organizations’ Alma-Ata declaration of 1978 outlined a global strategy for ‘health for all by 2000’ through the primary health care system. WHO estimated that the implementation of this strategy requires annual per capita health expenditure of US $15 in 1981 prices for most of the developing countries. Since per capita public spending on health is currently only US $2.3, there is an annual resource gap of US $50 billion for all the developing countries. Even if developing countries could fund as much as 50% of this amount, they would have to seek external funding of about seven times the present level of international transfers. The growth of per capita domestic public spending on health, therefore, is not anywhere near the levels required to meet the goals of the global strategy. [WHO (1981c), World Bank (1980b)]. Because of this reason and others (discussed in Chapter VI), various researchers and policy makers suggest alternative method of generating funds to finance health services (Jimanez 1986a, 1986b). Among the suggested measures are; user fees, development of suitable insurance system etc. It is in this context that the present study aims at examining the issue of financing the health services in developing countries in general and AP in particular. The study aims at reviewing the traditional arguments in favor of government intervention
- 49. 34 and the current arguments for introducing the user fees for health services by examining the resource allocation pattern in the state of AP. In order to examine the feasibility and its impact of introducing the user fees, the study aims at conducting a field study on government-run hospitals in AP. 2.3. Conclusion The purpose of this chapter was to review the past studies on hospital performance and efficiency with a view to identify the broad issues that were addressed in the literature. From the above review, we notice that earlier researchers have dealt with the issues of case mix, quality heterogeneity, uncertainty, and input price variations while analyzing hospital performance and efficiency. However, most of the above studies were for developed countries and we could find only a few studies for underdeveloped countries. Further, even the few studies that were done for developing countries could not address the above issues. One of the reasons could be non-availability of appropriate and detailed data. The next chapter tries to look into the alternative approaches that were used for evaluating hospital performance and efficiency. The objective is to identify an appropriate methodology, which can be used for evaluating hospital services in the state of Andhra Pradesh.
- 50. 35 CHAPTER - III HOSPITAL PERFORMANCE AND EFFICIENCY: ALTERNATIVE APPROACHES 3.0. Introduction This chapter gives a brief outline of different approaches that are used to measure hospital efficiency. We discuss basically four approaches, viz., cost accounting, ratio analysis, combined utilization and productivity (CUP) analysis and statistical production and cost functions that are used to measure hospital performance/efficiency. These approaches have varying data requirements. The first one uses the accounting information of different hospitals and analyzes the hospital service records in order to examine hospital costs and performance. It can be applied usefully to a single hospital and involves a labor-intensive detailed examination of hospital accounts, staffing pattern, and admissions. The second and third approaches are relatively simple and less data demanding. They use hospital census data on the number of beds available in the hospital, and number of inpatients and outpatients treated. On the other hand, the fourth approach requires detailed data on hospital functioning and could be used for evaluating the efficiency of many hospitals at a time. This method, based on micro economic theory, provides insight into cost issues- the relation between marginal and average cost the degree to which hospitals exhibit economies of scale and scope. The first two approaches do not focus on these aspects. 3.1. Cost accounting method The accounting method requires more detailed information on the components of hospital cost. Barnum and Kutzin (1993) have grouped cost accounting studies into two categories. The first category consists of studies which use detailed cost information in a ‘step down procedure’ to describe aggregate costs across departments and functions. The second category makes a less detailed estimate of hospital average cost based on aggregate central ministry information or hospital statistics and accounting records. 3.1.1. Step down procedure Step down cost accounting is a dis-aggregate method of analyzing the cost associated with specific hospital outputs. It is based on the scrutiny of the
- 51. 36 hospital production process to enable the best assignment of costs to outputs to which they are related. Total hospital expenditure is apportioned to specific departments (cost centers), and criteria such as ‘time use’ are employed to distribute all costs (including overhead and cost of intermediate outputs) to final service categories. Steps involved in a step down cost analysis are given below: The first step in the estimation process is to get a complete picture of total hospital recurrent cost. This means combining hospital line item expenditure (typical line items are salaries, drugs, other supplies, public utilities etc.) with data on resources used that do not appear in the hospitals budget or financial statement. Next step is to attribute line item costs to cost centers in specific hospital departments. The cost centers may differ across studies. Three typical cost centers used in Barnum and Kutzin (1993) are overhead, intermediate and final costs. The services produced by overhead cost centers are consumed by other departments (cost centers) of the hospital, not directly by patients. The cost of administration, housekeeping, and maintenance are included in this. The intermediate cost-centers generate services, which are not only consumed by various departments but also by patients directly. Examples of such costs are x- ray, laboratory tests, operation theatre etc. The final cost centers provide the services, which are directly consumed by the patients, not by other departments. These include outpatients and inpatients. Some studies dis-aggregate these services among different departments, e.g., medicine, surgery, orthopedics etc. However, there are data related problems associated with this step down cost analysis. First, cost data may not be directly available for individual hospitals. Second, multiple sources of budgeting, (e.g., central as opposed to district level) and assorted ways of making payments for different line items (e.g., the salaries of physicians and nurses may be paid by health ministry, other salaries may be paid by district authorities or by the hospital) make the reconstruction of actual expenditure tedious. Third, cost information may be available only on aggregate basis for the hospital. The advantage, however, is the need to reconstruct hospital cost data from multiple sources provides some insight into the problems of resource allocation that confront hospital managers. 3.1.2. Other accounting methods An alternative to step down method is to use aggregate data either for individual hospital or for a group of hospitals. The method relates cost and service
- 52. 37 information to time and institutional and geographical coverage. Studies based on this method require less time and less detailed information in comparison to step down procedure (Barnum and Kutzin 1993). The above procedures provide data on average cost of different services rendered by the hospitals. Though the average cost data are useful for assessing hospital performance, they are not sufficient for reaching definitive conclusions regarding hospital efficiency. Further, the above cost studies underlie the following assumptions: (a) The quality of service is uniform across facilities so that the cost per an equivalent unit of output could be compared. (b) The clinical composition of patients, i.e., case mix is the same at each facility. It must be emphasized that without an understanding of quality of service and case mix, the efficiency implications of variations in average cost can not be properly interpreted. For example, high average cost may reflect high quality, but low efficiency. On the other hand, low average cost may be a reflection of low quality and thereby low efficiency. Thus, in the absence of information on quality and case mix, it is difficult to draw meaningful conclusions from cost accounting studies. 3.2. Ratio analysis Another method that requires less detailed data and yet provides useful information on hospital performance is the ratio analysis. Three inter-related indicators of hospital services called average length of stay (ALS), bed occupancy rate (BOR), and bed turnover rate (BTR) are used for this purpose. These indicators are called performance indicators and are computed based on data from three basic hospital censuses, namely, bed capacity (B) of the hospital, cumulative inpatient days (IP) during a specified time interval, year in our case, and admissions (A) during the same time interval. Based on this data, the performance indicators are defined as below: (a) Bed turnover rate or flow, BTR = B A (b) Bed occupancy rate, BOR = 100 intervalTimeB IP
- 53. 38 (c) Average length of stay, ALS = A IP It is to be noted that bed capacity is in the denominator of two of these three indicators, implying that bed capacity is a global surrogate for inputs. As the staffing pattern, budget etc., are broadly linked to the bed strength through norms for provision of staff and budget, it is not inappropriate to use bed capacity as a surrogate of inputs. In addition, it is to be noted that these three indicators are interdependent and based on all the three hospital censuses (B, IP, and A). Knowledge of any two can give the value for the third 4. Hence, for the purpose of measurement of performance, any two indicators, say, BOR and BTR are sufficient. 3.3. Combined utilization and productivity analysis Lasso (1986) argued that the use of BOR, BTR and ALS individually do not provide adequate information and sometimes may give misleading results on hospital performance. The bed occupancy and turn over rates (BOR and BTR) may vary from country to country and between hospitals at different levels, namely primary (low), secondary (middle) and tertiary (high). Bed occupancy rate decreases as the level of hospital decreases. Further, a low BOR increases the average cost of services. Even if hospital inputs are being used with technical efficiency, low bed occupancy implies economic inefficiency. On the other hand, a high BOR does not necessarily indicate better hospital performance. A high BOR brings down the average cost but may be associated with low quality. Thus, the implication of high BOR for average cost and hospital efficiency is ambiguous without information on other service indicators. A high BOR with modest ALS may reflect an efficient situation (i.e. BTR is also high) or an inefficient situation when a large proportion of beds are occupied due to high ALS. Thus, explanatory power of a single indicator becomes clearer when information on other indicators is also available and used for the purpose (Barnum and Kutzin, 1993). Therefore, instead of using these indicators individually for measuring hospital performance, he developed a graphical technique for using the indicators 4 Annual bed occupancy measures the percentage of total available beds that are occupied by patients in a year and is calculated by the formula, BOR = IP / (B × 365). Annual turnover rate is calculated by the formula, BTR = A / B. Average length of stay is determined by ALS = IP / A. By simple manipulation we get, ALS = (BOR × 365) / BTR. Thus, once BOR and BTR are known, we can calculate ALS easily.
- 54. 39 together5. Implementation of CUP analysis involves three steps. First, basic data on hospital censuses, namely inpatient days, admissions, and number of available beds need to be collected. Second, performance indicators such as bed occupancy rate (BOR), bed turnover rate (BTR), and average length of stay (ALS) are computed. Third, BOR and BTR are plotted on a chart, where average bed occupancy (also referred to as utilization) per year and the average annual number of discharges per bed (i.e., BTR, also known as productivity) are shown on the x- and y-axis respectively. The chart is then divided into four sectors (regions) using two intersecting lines drawn at the mean values of BOR and BTR, which in turn identify the mean value of ALS. Each of the regions thus obtained has the following features: Sector I (lower left) indicates relatively low levels of bed occupancy and productivity, the least desirable situation. Sector II (upper left) indicates relatively low level of bed occupancy, but high productivity and short hospital stay. Sector III (upper right) indicates high level of bed occupancy and productivity, the most desirable situation. Sector IV (lower right) indicates relatively high level of bed occupancy, but low productivity and long hospital stays (Lasso 1986). It is to be noted that CUP analysis can give a broad picture about the performance of hospitals. This serves as an operational/benchmark tool for the hospital administrators in order to make policy decisions. The policy maker can identify the low performing hospitals on the basis of the performance indicators for further investigation on the causes of their low performance. It is obvious that hospital performance depends very much on the quality and clinical composition of cases treated. Comparison of different hospitals about their performance is therefore problematic. CUP analysis can partly mitigate this problem by grouping the hospitals into different categories. Apart from performance indicators, there are other aspects of a hospital such as multi-product nature, organizational structure, economies of scale and scope, which are of prime importance to the researchers as well as policy makers. These could not be analyzed through any of the three techniques just described. Such issues can better be addressed using multi-product cost functions. 5 The technique of using these indicators together is called combined utilization and productivity (CUP) analysis. Lasso (1986) defines BOR as utilization and BTR as the productivity of a hospital. When these two indicators are plotted together in a chart, it shows the productivity as well as the utilization levels of the hospital under study.
- 55. 40 3.4. Statistical/econometric approaches Health economists have used statistical and econometric approaches for various purposes, namely to evaluate the efficiency of hospitals, examine the determinants of hospital costs, address the issues relating to economies of scale and scope, and to find out the relationship between marginal and average cost. These approaches are generally based on micro economic theory of production and cost. In micro economic theory of production, we come across two concepts of efficiency, namely technical and economic efficiency. 3.4.1. Technical and economic efficiency A production process is said to be technically efficient if there is no way to produce more output with the same inputs or to produce same output with fewer inputs (Varian 1987, p.9). In the context of medical care, a process is technically efficient if production inputs (labor, drugs, equipment) are combined in a way that yields the maximum feasible output (outpatient visits, hospitalizations etc.). In theory of the firm terminology, it means that any point on the production possibility frontier is technically efficient. A production process will be technically efficient only in the presence of scale effects i.e., when economies of scale exists. In other words, diseconomies of scale may arise due to managerial, technological and socio-economic inefficiencies (Sherman 1981). Technical inefficiency in case of hospitals may arise due to inefficiency in the use of staff, supplies and equipment (Barnum and Kutzin 1993). Economic efficiency extends the concept of technical efficiency to take into account the relative costs of alternative inputs as well. A production process is called economically efficient, if inputs are combined to produce given level of output at minimum cost. Wyszewianski et al (1987) apply this concept in relative terms, arguing that efficiency improvement requires reducing the total cost of inputs used to produce a given output. In other words, if the total cost of inputs used by process X to produce a given level of output is less than the cost of inputs required by another process Y to produce the same output, then the process X is said to be more economically efficient than the process Y. The optimal input combination is given by the tangency point between an isoquant and an iso-cost (price) line. Thus, economic efficiency describes a specific subset of technically efficient choices. In fact, for a given set of input prices and output levels, there can be only one economically efficient combination of inputs,
- 56. 41 although many technically efficient alternatives might be possible. The statistical approaches make use of the production function in order to estimate (or otherwise characterize) the input-output relationships in health care sector. The shape of the production function is usually specified a-priori, on the basis of economic theory and practical considerations of simplicity and plausibility. Often some experimentation with alternative functional forms is required to obtain a ‘best’ fit or a suitable match between the data and the requirements of microeconomic theory. Similar procedure is followed for cost function estimation. Often cost function approaches are used, in lieu of a production function to deal with multiple outputs and inputs. Thus, production function is used to deal explicitly with multiple inputs (and single output), whereas cost function deals with multiple outputs and inputs, the latter considered implicitly (Sherman 1981). For instance, let us assume that the production function is specified as: )X...X,XF(Y m21 (3.1) where m21 ...X,X,X represent the quantity/value of each of m inputs and Y the corresponding output. The functional form F is assumed to be known and each output value, Y, is maximal in order that equation (3.1) represents a production function. Given this situation, we may distinguish the observed input amounts for some firm by indexing them as mj2j1j X...,,X,X so that ijX represents the amount of th i input used by firm j. Using these inputs, the th j firm produces the output jY which satisfies: jj YˆY0 (3.2) where, )X...,,X,X(FYˆ mj2j1jj (3.3)
- 57. 42 In other words, jYˆ is the (maximal) output, which would have been achieved if the firm had used these inputs efficiently. Equation (3.2) can be re-written in the ratio form: 1 Yˆ Y 0 j j (3.4) with an efficiency ratio of unity being achieved only if production process is fully efficient. Failure to achieve a ratio of unity would then represents what might be termed as ‘managerial or technical inefficiency’ irrespective of the prices paid for these inputs and without regard to the prices received for the output, provided that all prices are positive. Of course, a management, which achieves jYˆ but paid excessive prices for the inputs, would simultaneously be economically inefficient. This represents price inefficiency. Closely related to the concept of price inefficiency is ‘allocative inefficiency’. Such allocative inefficiency can arise, even if the prices paid are minimal, when the firm does not respond to the opportunities the market allows for substituting between factors in order to achieve a given (maximal) output level jYˆ . In other words, the wrong (inefficient) mix of inputs is used to achieve this output. 3.4.2. Evaluation of hospital efficiency Regression analysis approaches to estimating input-output relationship are widely used, but these are subject to serious limitations when applied to non- profit oriented sectors and their efficiency evaluations. Such limitations arise from the assumption that the relationship estimated by the regression techniques reflects efficient input-output relationships. This could only be true if the observations were all generated from the hospitals that are known to be efficient. Such an assumption is questionable, to say the least, in a sector like hospital services where, clearly, pure competition in the sense of market price economics may not hold. Another set of problems which affects most of the existing regression based studies of health care organizations is, the need to aggregate multiple outputs into a single output measure. However, we will not be dealing with the studies,
- 58. 43 which convert multiple outputs to single output. Hence, we confine to statistical approaches which deal with multiple output situations such as (a) simultaneous equation models including the recursive type and (b) single equation models, which utilize flexible functional forms to deal explicitly with multiple outputs and inputs together. The first of these alternatives relates to causal identification via the way equations are modified relative to each other. The problem of causal identification, which occurs in simultaneous equation models, has not been addressed in most of hospital services literature, except Feldstein (1967). The translog function, a representative of the flexible functional forms, has been receiving increased attention in the study of health services efficiency. We shall examine the usefulness of this and other functional forms in analyzing hospital efficiency. Our treatment of these statistical/econometric approaches for evaluating hospital efficiency will proceed as follows. In the following section, we will review the translog function as applied to the health sector. This is our primary interest because, (a) it explicitly allows for simultaneous evaluation of multiple outputs and inputs which is an important attribute for hospital evaluation, and (b) it does not require that specific functional form be specified a-priori which is a critical feature of hospitals since the production and cost function of hospitals are not known with any reasonable degree of precision. Next we consider the set of hospital production and cost functions with emphasis on the study by Feldstein (1967). 3.4.3. Production and cost functions Production functions We want to consider certain background materials that are necessary for understanding flexible functional forms. The easiest way is to consider the case of an ordinary Cobb-Douglas function involving one output (Y) and two inputs ( 21 XandX ). The function has the form: β 2 α 1 XXγY (3.5) Where 0βα, and >0 are constants and represent output elasticities w.r.t.
- 59. 44 factor inputs 21 XandX respectively. The production process exhibits increasing, constant or decreasing returns to scale when βα 1 respectively. The function (3.5) can also be written as: 21 XlnβXlnαln γYln (3.6) Equation (3.6) is the usual form for fitting the Cobb-Douglas function by the method of least squares to determine the values of βandαγ, . Statistical tests of significance can be applied to determine, among other things, whether βα 1 applies. The above functional form can be extended to the case of more than two inputs, say m, as below: i m 1i i Xlnαln γlnY (3.7) It is well known that the Cobb-Douglas production function implied in both the equations (3.6) and (3.7) exhibits unitary elasticity of substitution, which is considered an unrealistic property. To avoid this, other functional forms like constant elasticity of substitution (CES), variable elasticity of substitution (VES) and a host of others are proposed in the empirical literature. Empirical analysis usually involves in specifying alternative functional forms, estimating them with the data on hand and choosing the ‘best’ functional form, keeping the statistical and economic criteria in mind. Christensen, Jorgenson and Lau (1973) have proposed a general functional form known as Transcendental logarithmic form, Translog for short, to represent a second order approximation of Taylor’s series expansion for any production function. This functional form has become instantaneously very popular and extended to utility, cost and profit functions as well. The advantage of this flexible functional form is that the neoclassical properties of a production function can easily be imposed as parametric restrictions and their empirical validity can be tested using data. Further, translog form allows the implementation and empirical testing of a priori restrictions of additivity/separability of factor inputs. The general translog production function can be specified as below: j m 1i m 1j m 1i XlniXlnijβ)21(iXlniα0αYln (3.8)
- 60. 45 It may be noted that (3.8) can be estimated using single equation or systems approach. The latter is necessary when we are interested in estimating the output supply and factor demand equations corresponding to the translog production function. Advanced econometric techniques like Zellner’s seemingly unrelated regression estimation (SURE) are used in this case. Production functions such as (3.8) along with neoclassical restrictions imposed on the parameters are sometimes referred to as ‘structural’ models. A number of empirical studies, as already listed in Chapter II, have estimated such structural models for health sector in developed countries. Useful discussions about efficiency (technical and economic) are carried out using the estimated production function. Researchers have worked out the expressions for returns to scale, factor demand elasticities and Allen-Uzawa partial elasticities of substitution underlying (3.8). An additional advantage of (3.8) is that the Cobb-Douglas function given in (3.7) as well as few other functional forms such as log-quadratic can be obtained as special cases by restricting the ij parameters appropriately. It is clear that (3.8) deals with a single output (Y) and several inputs (X), m in number here. Unfortunately, the multi-product generalization of the production function in (3.8) is not easy to conceive or specify. However, as we shall show shortly, the single and multi-product parallels of translog cost function are relatively easier to specify. Recursive models: Feldstein (1967) has proposed a recursive model to represent the production behavior of hospital services in Great Britain. The model has the ability to handle both multiple outputs and inputs, albeit in somewhat a special way. The recursive model by Feldstein (1967) is as follows: (i) β)exp(p'MBD 21 αα (ii) 543 ααα DˆMBW (iii) β)exp(p'MBR 76 αα (3.9) (iv) β)exp(p'RˆBN 98 αα (v) β)exp(p'RˆBH 1110 αα where D: drugs and dressing expenditures, B: number of beds, M: number of
- 61. 46 medical staff, W: patient outputs in physical units, R: proportion of occupancy, N: nursing expenditures, H: housekeeping expenditures and the bracketed term is the vector of case mix proportions weighted by a vector . The model has the following assumptions: (a) No hospital output is constrained by exogenously determined levels of nursing (N) and house keeping (H) expenditures. Rather, these variables are determined by the hospital, based on its level of operation. (b) Output, measured in number of patients treated, depends only on number of beds (B), medical staff (M), drugs and dressings (D) and case mix. (c) Beds (B), medical staff (M) and the vector of proportion of patient case types (p) are exogenously determined. It can be seen that in the model, each equation specifies a Cobb-Douglas type of function of the inputs and an exponential component to handle case mix variations, e.g. proportion of severe vs. normal cases and ambulatory or outpatient treatments, etc. Closure inspection of (3.9) shows that the model is recursive with two sub-systems, namely equations (i) and (ii); and equations (iii), (iv) and (v). In the first sub-system, the drugs (D), which are the outputs in equation (i) becomes input to equation (ii), but do not appear anywhere else in the system. The second sub-system has proportion of occupancy (R) as an output in equation (iii) and as an input to equations (iv) and (v). The two sub- systems are tied together by common (exogenous) input factor beds (B). Since the above model is recursive, ordinary least squares estimation was used to estimate it in the following log-linear form: k 1j jj21 βpMlnαBlnαDln DˆlnαMlnαBlnαln W 543 k 1j jj76 βpMlnαBlnαRln (3.10)
- 62. 47 k 1j jj98 βpRˆlnαBlnαNln k 1j jj1110 βpRˆlnαBlnαHln Feldstein (1967) compared the results of this recursive model with a similar set of single equation regressions, also in Cobb-Douglas form. The comparison favored the recursive model. The results also indicate, rather weakly, the presence of decreasing returns to scale in hospital services. Cost functions For a variety of reasons, it is easier to study the aspect returns to scale using cost functions (Cowing, Holtman and Powers, 1983) and this in any case, is the way most of the studies in health economics have approached this topic. In addition, many hospital studies are primarily concerned with cost behavior rather than input-output relationships addressed directly by production functions. A simple cost function may be represented as: p)(Y;CC (3.11) where total cost (C) is a function of the output vector Y and the vector of input prices (p). The cost function has the ability to consider the effect of multiple outputs on total cost. On the other hand, it deals with inputs only implicitly although, provided technical efficiency can be assumed, modern theories of Shephard-Samuelson duality permit one to move between cost and production functions to remedy this deficiency. In any case, ‘p’ serves as a vector of parameters with (under optimising behavior) the vector of outputs (Y) resulting from input mixes and magnitudes selected under whatever technology (Baumol, Panzar and Willig 1982). Shephard [reproduced in Baumol , Panzar and Willig (1982)] used equation (3.11) for his duality relations between cost and production functions, which inter alia makes it possible to move in a relatively easy manner between cost and production functions. Shephard lemma (see for example, Bothwell and Cooley 1978) states:
- 63. 48 j j X p p)C(Y; (3.12) where jX represents the optimal amount of the th j input. Therefore, at any specified level of output and prices, the optimal inputs, jX (j=1,2,…,n), are directly determinable from the cost function. The duality theory, which relates outputs to input in the above-indicated manner, assumes that ‘technical efficiency’ has already been attained in both cost and production. With this assumption, it is possible to move from cost to production functions and vice versa, so that the analyst is free to use either of the two approaches. A specific case of equation (3.11) is the translog cost function, which can be represented as: m 1i m 1i m 1j jYlniYlnδ 2 1n 1j jplnjβilnYiα0αCln ij m 1j n 1k kplnjYlnjkρ 2 1n 1j n 1k kplnjplnjkγ 2 1 (3.13) where, C: total cost, jp : unit price of th j input and iY : amount of th i output. A number of studies have used the translog cost function for measuring hospital efficiency and other related aspects. In addition to the pure translog cost function given above, some researchers have addressed the topic of efficiency as well as economies of scale and scope through hybrid functional forms. In these functional forms, the independent variables, particularly the price variables, enter the equation in logarithmic form. The output and other variables appear in actual scale without the logarithm. The translog hybrid cost function can be specified as:
- 64. 49 m 1i m 1i m 1j jYiYδ 2 1n 1j jplnjβiYiα0αCln ij m 1j n 1k kplnjYjkρ 2 1n 1j n 1k kplnjplnjkγ 2 1 (3.14) The definition of ‘Y’, ‘p’ and ‘C’ remains the same as above. In the following section, we give a brief review of some of the selected functional forms that are widely used in hospital cost function studies. Hospital Cost Functions Feldstein (1967) approached the topic ‘returns to scale’ from the standpoint of cost functions. He began by examining the cost per case as a function of number of beds (B), which is a measure of scale. Two simple average cost functions were estimated with number of beds (B) in linear and quadratic form separately as below: BβαAC (3.15) 2 21 BβBβαAC (3.16) where AC is the average cost per patient, and are parameters to be estimated. Though the goodness of fit was poor using the above functional forms, the results indicated increasing average cost in certain range of hospital size (number of beds). Average cost regression was re-run with a case mix variable reflecting the nine specialties (e.g., pediatrics, general surgery, general medicine etc.) added: 9 1i ii 2 21 pδBβBβαAC (3.17) where p i ( ...,9)1,2,i is the proportion of case mix types in relation to all cases treated in a hospital. In addition to substantial increase in explanatory power, the case mix variable became significant in the regression. Likewise, inclusion of
- 65. 50 case flow rate variable (average number of cases treated per bed year, F) also improved the results further. 2 9 1i ii 2 21 FρFδpγBβBβαAC (3.18) In equation (3.18), the average cost per patient was found to decrease as hospital bed size increases implying increasing returns to scale. In contrast, as noted earlier, the production function estimates reflected slight decreasing returns to scale for hospitals of certain size. Different authors used similar cost functions6. These studies on hospital cost concentrated largely upon determining the shape of the average cost curve, effects of case mix on hospital cost, and optimal hospital size by using cross sectional data on various hospitals. In addition, these studies also focus on estimating the possibility of scale economies, which implied increased efficiency for larger hospitals. However, most of these studies did not pay adequate attention to the multiple output, or input price issues as discussed in the previous chapter. Although some of these studies (particularly in 70’s) gave more attention towards the case mix problems, they failed to link the empirical models used to the theoretical concept of multiple-output firm and the related cost structure. Thus, their work is considered “less regorous, even rather ad-hoc at times, than might be desired, a weakness which severely limits its interpretation and usefulness for policy analysis” (Cowing et al 1983). More recent studies on hospital cost use theoretically consistent models of hospital cost. These studies specify the total cost to be a function of multiple categories of outputs and inputs. In addition, various measures of capital stock (e.g. number of beds, number of admitting physicians etc.), and case mix are also included in these models. These studies have used either translog or hybrid functional forms in order to estimate the relationship between the cost and output. One of the early applications of translog cost function was by Cowing and Holtman (1983). The authors used a functional form similar to that given in equation (3.13). They used five outputs (emergency room care, medical surgical care, pediatric care, 6 For example, see Carr and Feldstein (1967), Newhouse (1970), Bays (1977), Pauly (1978), Lave and Lave (1970), and Rafferty (1971).
- 66. 51 maternity care and other care), prices of six inputs (nursing labor, auxiliary labor, professional labor, general labor, and materials and supplies), and two measures of fixed variables (number of admitting physicians and fixed capital) in their model. The model is specified as follows: m 1i m 1i m 1j jYlniYlnijδ 2 1n 1j jplnjβilnYiα0αCln SlnlnYψ m 1i n 1j jplniYlnijρ n 1j n 1k kplnjplnjkγ 2 1 i m 1i i DlnSlnSlnplnπ sd n 1j jj 2 sss S)(lnSln 2 ddd D)(logθDlogθ (3.19) where, C: variable cost, D: number of admitting physicians (assumed to be fixed in the short run), S: measure of fixed capital stock. Using Shephard lemma, the authors derived the cost share equations for all the variable inputs. Thus, the actual model has seven equations, namely six share equations and the cost function itself. By using cross-section data from 138 non-federal, short-term, general-care hospitals in USA, the above model with six equations7 was estimated by using the method of maximum likelihood. Probably, the most frequently used functional form in recent times is the hybrid cost function originally proposed by Grannemann et al (1986). The functional form is given below: Z))R,CM,D,exp(f(Y,pAC iβ i n 1i (3.20) or 7 Since cost shares add-up to unity, out of the six share equations only five are linearly independent and can be estimated. The parameters of the sixth equation can be obtained using all other parameters.
- 67. 52 Z)R,CM,D,f(Y,plnβAlnCln i n 1i i (3.21) where A: vector of various factors that are assumed to affect the level of cost, but not the shape of the cost function with respect to outputs, pj: price of jth production input, Y: vector of hospital primary outputs (Y1, Y2, Y3 are inpatient days by type, Y4 is outpatient visits, Y5 emergency department visits), D: vector of inpatient discharges by type, CM: matrix of case mix variables, R: vector of various sources of revenue of the hospital, Z: vector of other miscellaneous outputs produced by the hospital. The outputs enter the equation in linear, quadratic as well as cubic forms and other variables such as CM and R enter with interaction terms. Functional forms similar to Grannemann et al. (1986) have been used in Bitran and Dunlop (1989), Vita (1990), Barnum and Kutzin (1993) and various other studies. Mixed models The above models are primarily used for examining the determinants of hospital costs, estimating the extent of scale and scope of hospitals. However, in studies where the primary focus is to examine the efficiency of the hospital production process, use somewhat different approach. These studies, in addition to the use of cost functions, use the production function approach in order to calculate the efficiency index. Goldman and Grossman (1983) have used this approach for estimating the inefficiency of community health centers in the U.S. Similar approach was followed by Wouters (1990) in her study of cost and efficiency of a sample of 42 private and public hospitals in Nigeria. Technical efficiency was estimated using a production function. Measures of marginal product of health workers are also obtained. The following production function was used: LHWlnaHHWlnaDRUGNUM86lna λ 1)(IN aalnV 432 λ 10 BEDSDUMMY λ 1)(BEDS aBEDSDUMMYa λ 65 (3.22)
- 68. 53 where is a parameter of the Box-Cox transformation. In the above production function, (logarithm of) the number of outpatient visits (V) is the dependent variable. Independent variables are, the number of inpatient admissions (IN), the number of patients receiving drugs (DRUGNUM86), the proportion of high and low level health workers (HHW, LHW) and the presence of number of beds (BEDSDUMMY and BEDS). To examine whether the cost minimization was taking place, estimates of marginal product of high and low level worker were compared with relative category of labor in private and public facilities. She found that pubic facilities employed too many low-level workers compared to private hospitals. In contrast, private providers were found to be using a near optimal (cost minimizing) mix of high and low level of health workers. The estimated cost function is as follows: INDEXlnbDRUGPCT86lnb )1IN( bVlnbbRCOSTln 43210 BEDSDUMMYbln WAGLHWbln WAGHHWb 765 BEDSDUMMY )1BEDS( b8 (3.23) where RCOST: recurrent expenditure of the facility, DRGPCT86: proportion of patients obtaining drugs, WAGHHW and WAGLHW: wages of high and low level health workers and INDEX: efficiency index defined as, 1 WAGE WAGE MP MP INDEX HHW LHW HHW LHW (3.24) The other variables are as defined in equation (3.23). The equation (3.23) was estimated using OLS method. Thus, in this chapter, we have reviewed the alternative approaches that were used in the literature for evaluation of hospital performance and efficiency. We notice that the method of combined utilization and productivity (CUP) was used for evaluating hospital performance and statistical cost functions for hospital
- 69. 54 (economic/allocative) efficiency. CUP analysis is a managerial technique that could be used for quick identification of low performing hospitals. This method does not require detailed data on the hospital production process and depends mostly on hospital census data such as inpatient days, outpatient visits, admissions, beds etc. However this method does not take certain critical issues such as economies of scale and scope, marginal and average costs, elasticity of substitution etc., which are of prime importance to the economists and policy makers. These issues are well handled by the statistical cost functions. We therefore, attempt to use the CUP analysis for evaluating hospital performance and cost functions for efficiency in Andhra Pradesh. The specific objectives of this study are listed below: 3.5. Specific objectives of the study: 1. Evaluate the performance and efficiency of secondary level hospitals in the state of Andhra Pradesh. This is attempted through: (a) Combined utilization and productivity (CUP) analysis for evaluating hospital performance. (b) Cost functions for evaluating hospital efficiency. 2. Examining the issue of health financing in developing countries in general and AP in particular, with a view to suggest some policy measures in this regard. This is done through: (a) Review the methods of financing the health sector as suggested by various international agencies. (b) Assessing the resource allocation pattern in the health sector in AP. (c) Examine the feasibility of user fees as a source of financing in Andhra Pradesh district hospitals using a field study. In the next chapter, we take up the empirical analysis of evaluating hospital performance in the state of Andhra Pradesh using CUP analysis.