eMJA Casemix Supplement: DRG cost weights

Contents | Search | eMJA Home

eMJA

Casemix: moving forward

DRG cost weights -- getting it right

Peter D Phelan, Richard Tate, Fiona Webster and Ric P Marshall

MJA 1998; 169: S36-S38

Synopsis - Introduction - Costing methods - Cost model approach - Shortcomings of cost modelling - Patient cost data approach - Value for clinicians - Victorian cost weights - A simplified patient costing system - Conclusion - Reference - Authors' details

Synopsis

Inadequate cost weights are a major problem in casemix funding systems.

Clinicians should understand the basis for the cost weights underpinning the hospital payment system in their State and their own hospital.

Clinician managers need valid patient costing data if they are to benchmark and improve cost-effectiveness while maintaining and enhancing quality.

The cost model approach for determining cost weights has inherent limitations, and, the alternative, detailed patient costing, requires efficient hospital information technology systems. A simplified approach to patient costs, which uses existing hospital data systems, may be useful for smaller hospitals.

A better classification system and funding formulas incorporating reliable cost weights derived from patient costing should overcome many of the deficiencies in the current casemix payments systems.

Introduction
One of the major problems associated with introducing DRG weighted funding for acute-care hospitals in Australia has been the inadequacy of the cost weights used in some States. This stems firstly from a lack of reliable patient cost data from a sufficiently broad spectrum of hospitals to modify the DRG classification, and secondly from clinicians' inadequate understanding of how cost weights are developed. The latter has insulated hospital administrations and funding agencies from demands for improved costing systems. Most complaints by clinicians about DRG classification or funding deficiencies would be resolved if there were better patient costing systems in hospitals. Better costing systems would also be invaluable to clinician managers. Clinicians should be insisting that their hospitals do better in this area.
Costing methods
There are two accepted approaches to allocating costs to patients in a DRG.
The Yale cost model¹ is a "top down" method in which assignment of costs is based on data from the hospital's central accounts. Its major limitation is that it is dependent on the accuracy of these data. It was the basis for the national cost weights developed by the Federal Government in the early 1990s (and subsequently modified very superficially). Further national cost weights are currently being developed; they also appear to be based on cost modelling.
Patient cost data is a "bottom up" method in which the actual costs of individual patient episodes of care are recorded, including the resources consumed and an appropriate share of the overheads. The DRG cost weight is the average "true" cost of all patients within that DRG. Ideally, this method requires a computerised "feeder system" to capture data on all aspects of a patient care episode. This approach has been used in Victoria for inpatient casemix funding.
Cost model approach
The cost model approach does not require departmental feeder systems and thus avoids the need for investment in information technology to record patient costs. The basic information comes from the hospital's central accounts. First, a decision is made as to what fraction of the hospital's overall expenditure is consumed by inpatients -- the IFRAC or inpatient fraction. This is then applied to the cost centres, such as wards, medical salaries, operating room, pharmacy, radiology and pathology, social work and other allied health services.
Patient costs are distributed according to predetermined service weights, based on the relative costs of nursing, pathology, radiology and so on, over all the DRGs. These weights, in turn, are derived from previous studies (such as the one for developing the Yale cost model). For example, if the average nursing cost for patients in DRG 100 is taken as 1.00 unit, the average nursing cost for patients in DRG 150 could be calculated as a proportion of that -- for example, it may be 0.7. In this way, total nursing costs can be allocated to all patients once their DRG classification is determined. Overhead costs are generally allocated according to length of hospital stay.
The initial service weights were developed in Maryland (USA) and were based on amount charged, not what the service cost. There have been several Australian studies to develop Australian service weights. Their reliability varies according to whether clinicians' opinions or actual data were used.
Shortcomings of cost modelling
Cost modelling depends firstly on having reliable data (IFRAC), and secondly on the hospital having reasonable accounting structures and systems, with accurate data posted to the various cost centres. These criteria were not met by many hospitals in the original Australian national cost weight study. To what extent these problems will be overcome in the current national study is unknown.
Our experience indicates that simply accepting financial data supplied by hospitals, without a detailed on-site examination of their financial procedures and records, is likely to result in flawed data. This is of particular concern when the data used are not part of the hospital's regular review processes.
There are other problems with cost weights developed through cost modelling:
Their usefulness depends on the availability of reliable service weights. The reliability of Australian service weights has been questioned, because charges were used as a proxy for cost.

Cost modelling assumes that the same service weights apply across all hospitals irrespective of their patient mix. When no service weights are available, costs are distributed according to length of stay, which ignores differences in illness severity, cost or nursing dependency, for example, in the one DRG.

Cost weights provide "cost" information only at a DRG level. To refine the DRG classification system further, cost data are required at an ICD-9 or ICD-10 level. Until such information is available, it will not be possible to refine the classification system to take into account issues such as severity and complexity. To date, almost all classification changes in Australia have used length of stay as a proxy for cost; with ever decreasing lengths of stay, its reliability as a proxy diminishes.

Data available at a DRG level only, and established by predetermined cost weights, is limited in its usefulness to assist clinicians to modify their practice behaviour in line with economic realities.

Patient cost data approach
For this approach hospitals need a sophisticated computerised patient costing system. However, the costs of patient care can also be calculated with existing hospital data systems, as is described later in this article.
Organisational overhead costs: Patient costing starts by identifying organisational overhead costs and distributing these to each hospital department according to its share of the cost. Overheads for cleaning, staff numbers, patient activity or budget can be allocated on the basis of floor space. The distribution of overheads to departments can be either simple or complex, depending on the level of accuracy desired.
Departmental overhead costs are then attributed to inpatient and outpatient activity. Each cost centre must have information systems so all patient costs are traced to the patient receiving the service. The quality of the information systems and the accuracy of individual service costing in each cost centre determines the reliability of the final cost data.
Nursing staff costs: Nursing staff costs -- the single largest element in hospital expenditure -- provide a major challenge. A robust nurse dependency or care management system, which relates nursing hours to the degree of patient illness, is required; when this is not available, nursing costs either at a ward or unit level, or across the whole hospital, are distributed on the basis of length of stay (with its inherent limitations).
Department costs: Pathology and radiology departments can usually attribute tests to each patient, and provided this can be done for individual services, excellent cost data can be obtained. Pharmacy costs require the recording of pharmaceuticals dispensed to individual patients. If the hospital has only a ward imprest system, the tracking of cost data is less than optimal, as costs may have to be distributed on the basis of length of stay.
Operating room costs: These are usually based on staff time for each procedure recorded in the theatre register. In many hospitals, a major problem in allocating operating room costs has been surgical supplies, especially prostheses. A common practice is to distribute their cost in the same ratio as operating room time. This has led to substantial under costing of some DRGs grouping patients for whom expensive prostheses are required (eg, coronary artery stents). Their costs should be attributed to individual patients and not distributed across all patients using the operating room or radiology department (depending which cost centre purchases them). The advent of new and high-cost prostheses has highlighted the need for accurate tracking systems that allow costs to be allocated appropriately to different or new DRGs.
Medical staff costs: These present a problem as there is no simple system to allocate costs to individual patients, and doctors are unlikely to agree to keep a work diary. Therefore, medical staff costs are usually allocated to clinical units from staff rosters, and then distributed to patients (eg, according to length of stay in the unit).
Value for clinicians
It is important for clinicians to understand and contribute to the development of patient costing systems, and thus enhance the reliability of patient data. With the patient cost data approach, which is based on actual patient data, clinicans will better understand how cost weights are developed. If patient data collection is deficient in their hospital, they will appreciate how this affects cost weights. Access to reliable patient costing data allows benchmarking, which is essential if clinicians are to be effective managers and control costs, while maintaining and improving quality.
Victorian cost weights
Cost weights in Victoria have been developed from hospitals with computer based costing systems. There were five hospitals in the initial study and 15 in the 1996-97 study. Having individual patient data coded according to ICD-9 has made it possible to identify unexpected variations and outliers, and to go back to the source hospital for data verification, especially for low volume DRGs. Knowledge of the adequacy of each hospital's costing system has been important in achieving valid cost weights.
Variation in the cost weights from year to year has been partly due to the changes in the national classification system (AN-DRGs), to the impact of steadily improving hospital costing systems (by introducing new and improved feeder systems), to enhanced data collection, and to clinical variability within small volume DRGs from one year to the next.
There have also been actual changes in reported costs because of gradually reducing length of stay, increases in day surgery and changes in clinical management practice (eg, use of different drugs, operative procedures or prostheses). An annual update of cost weights is essential, if hospitals are to be confident about the cost weights underpinning casemix payment formulas. Once systems are in place, this can be achieved at reasonable cost.
A simplified patient costing system
Many hospitals do not have patient costing systems, because they cannot afford such a system or because their size does not justify the investment. A method has been developed that emulates the results of more expensive and sophisticated systems, but at a fraction of the cost. It is a database management system that draws together clinical and cost data at the patient level. This enables "bottom up"-based costing of patients which can then be summed to give overall costs and hence cost weights. Although this method does not yield the sophisticated data of the larger systems, clinicians often find themselves overwhelmed by data produced by larger systems, when in fact they would prefer fewer but more strategic and relevant data.
Conclusion
Cost modelling has inherent limitations, and the alternative -- detailed patient costing -- requires efficient hospital information technology systems. However, both use the same cost disaggregation systems for overheads and medical costs and, and in hospitals without nurse dependency systems, also for nursing costs. The suggested simplified approach to patient costs may be attractive for smaller hospitals, as it avoids considerable capital expenditure.
A better classification system and funding formulas incorporating reliable cost weights derived from "bottom up" patient costing systems should overcome many of the deficiencies identified by clinicians in the current casemix payments systems. It should better describe patients, particularly in teaching and specialist hospitals, and also smaller hospitals, each of which has a unique AN-DRG patient mix.
References

Fetter RB, Thompson JD, Mills RE. A system for cost and reimbursement control in hospitals. Yale J Biol Med 1976; 49: 123-136.

Authors' details
Faculty of Health Sciences, La Trobe University, Melbourne, VIC.
Peter D Phelan, MD, FRACP, MRACMA, Adjunct Professor.
Health Solutions Pty Ltd, Melbourne, VIC.
Richard Tate, BA(SocSci), GradDipAccount, Group Executive.
Healthwise Consulting, Mebourne, VIC.
Fiona Webster, BSc, MBA, GradDipApplPsych, AFCHSE, Health Consultant.
Acute Health Division, Department of Human Services, Melbourne, VIC.
Ric P Marshall, BA, DipPsych, GMQ, Manager, Information and Performance Evaluation.
Reprints will not be available from the authors.
Correspondence: Professor P D Phelan, Victorian Medical Postgraduate Foundation, PO Box 27, Parkville, VIC 3052.
E-mail: vmpfATvicnet.net.au

Go to the next article . . .
Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company.
Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>".
<URL: http://www.mja.com.au/> © 1998 Medical Journal of Australia.

Synopsis
	Inadequate cost weights are a major problem in casemix funding systems. Clinicians should understand the basis for the cost weights underpinning the hospital payment system in their State and their own hospital. Clinician managers need valid patient costing data if they are to benchmark and improve cost-effectiveness while maintaining and enhancing quality. The cost model approach for determining cost weights has inherent limitations, and, the alternative, detailed patient costing, requires efficient hospital information technology systems. A simplified approach to patient costs, which uses existing hospital data systems, may be useful for smaller hospitals. A better classification system and funding formulas incorporating reliable cost weights derived from patient costing should overcome many of the deficiencies in the current casemix payments systems.
Introduction
	One of the major problems associated with introducing DRG weighted funding for acute-care hospitals in Australia has been the inadequacy of the cost weights used in some States. This stems firstly from a lack of reliable patient cost data from a sufficiently broad spectrum of hospitals to modify the DRG classification, and secondly from clinicians' inadequate understanding of how cost weights are developed. The latter has insulated hospital administrations and funding agencies from demands for improved costing systems. Most complaints by clinicians about DRG classification or funding deficiencies would be resolved if there were better patient costing systems in hospitals. Better costing systems would also be invaluable to clinician managers. Clinicians should be insisting that their hospitals do better in this area.
Costing methods
	There are two accepted approaches to allocating costs to patients in a DRG. The Yale cost model¹ is a "top down" method in which assignment of costs is based on data from the hospital's central accounts. Its major limitation is that it is dependent on the accuracy of these data. It was the basis for the national cost weights developed by the Federal Government in the early 1990s (and subsequently modified very superficially). Further national cost weights are currently being developed; they also appear to be based on cost modelling. Patient cost data is a "bottom up" method in which the actual costs of individual patient episodes of care are recorded, including the resources consumed and an appropriate share of the overheads. The DRG cost weight is the average "true" cost of all patients within that DRG. Ideally, this method requires a computerised "feeder system" to capture data on all aspects of a patient care episode. This approach has been used in Victoria for inpatient casemix funding.
Cost model approach
	The cost model approach does not require departmental feeder systems and thus avoids the need for investment in information technology to record patient costs. The basic information comes from the hospital's central accounts. First, a decision is made as to what fraction of the hospital's overall expenditure is consumed by inpatients -- the IFRAC or inpatient fraction. This is then applied to the cost centres, such as wards, medical salaries, operating room, pharmacy, radiology and pathology, social work and other allied health services. Patient costs are distributed according to predetermined service weights, based on the relative costs of nursing, pathology, radiology and so on, over all the DRGs. These weights, in turn, are derived from previous studies (such as the one for developing the Yale cost model). For example, if the average nursing cost for patients in DRG 100 is taken as 1.00 unit, the average nursing cost for patients in DRG 150 could be calculated as a proportion of that -- for example, it may be 0.7. In this way, total nursing costs can be allocated to all patients once their DRG classification is determined. Overhead costs are generally allocated according to length of hospital stay. The initial service weights were developed in Maryland (USA) and were based on amount charged, not what the service cost. There have been several Australian studies to develop Australian service weights. Their reliability varies according to whether clinicians' opinions or actual data were used. Shortcomings of cost modelling Cost modelling depends firstly on having reliable data (IFRAC), and secondly on the hospital having reasonable accounting structures and systems, with accurate data posted to the various cost centres. These criteria were not met by many hospitals in the original Australian national cost weight study. To what extent these problems will be overcome in the current national study is unknown. Our experience indicates that simply accepting financial data supplied by hospitals, without a detailed on-site examination of their financial procedures and records, is likely to result in flawed data. This is of particular concern when the data used are not part of the hospital's regular review processes. There are other problems with cost weights developed through cost modelling: Their usefulness depends on the availability of reliable service weights. The reliability of Australian service weights has been questioned, because charges were used as a proxy for cost. Cost modelling assumes that the same service weights apply across all hospitals irrespective of their patient mix. When no service weights are available, costs are distributed according to length of stay, which ignores differences in illness severity, cost or nursing dependency, for example, in the one DRG. Cost weights provide "cost" information only at a DRG level. To refine the DRG classification system further, cost data are required at an ICD-9 or ICD-10 level. Until such information is available, it will not be possible to refine the classification system to take into account issues such as severity and complexity. To date, almost all classification changes in Australia have used length of stay as a proxy for cost; with ever decreasing lengths of stay, its reliability as a proxy diminishes. Data available at a DRG level only, and established by predetermined cost weights, is limited in its usefulness to assist clinicians to modify their practice behaviour in line with economic realities.
Patient cost data approach
	For this approach hospitals need a sophisticated computerised patient costing system. However, the costs of patient care can also be calculated with existing hospital data systems, as is described later in this article. Organisational overhead costs: Patient costing starts by identifying organisational overhead costs and distributing these to each hospital department according to its share of the cost. Overheads for cleaning, staff numbers, patient activity or budget can be allocated on the basis of floor space. The distribution of overheads to departments can be either simple or complex, depending on the level of accuracy desired. Departmental overhead costs are then attributed to inpatient and outpatient activity. Each cost centre must have information systems so all patient costs are traced to the patient receiving the service. The quality of the information systems and the accuracy of individual service costing in each cost centre determines the reliability of the final cost data. Nursing staff costs: Nursing staff costs -- the single largest element in hospital expenditure -- provide a major challenge. A robust nurse dependency or care management system, which relates nursing hours to the degree of patient illness, is required; when this is not available, nursing costs either at a ward or unit level, or across the whole hospital, are distributed on the basis of length of stay (with its inherent limitations). Department costs: Pathology and radiology departments can usually attribute tests to each patient, and provided this can be done for individual services, excellent cost data can be obtained. Pharmacy costs require the recording of pharmaceuticals dispensed to individual patients. If the hospital has only a ward imprest system, the tracking of cost data is less than optimal, as costs may have to be distributed on the basis of length of stay. Operating room costs: These are usually based on staff time for each procedure recorded in the theatre register. In many hospitals, a major problem in allocating operating room costs has been surgical supplies, especially prostheses. A common practice is to distribute their cost in the same ratio as operating room time. This has led to substantial under costing of some DRGs grouping patients for whom expensive prostheses are required (eg, coronary artery stents). Their costs should be attributed to individual patients and not distributed across all patients using the operating room or radiology department (depending which cost centre purchases them). The advent of new and high-cost prostheses has highlighted the need for accurate tracking systems that allow costs to be allocated appropriately to different or new DRGs. Medical staff costs: These present a problem as there is no simple system to allocate costs to individual patients, and doctors are unlikely to agree to keep a work diary. Therefore, medical staff costs are usually allocated to clinical units from staff rosters, and then distributed to patients (eg, according to length of stay in the unit). Value for clinicians It is important for clinicians to understand and contribute to the development of patient costing systems, and thus enhance the reliability of patient data. With the patient cost data approach, which is based on actual patient data, clinicans will better understand how cost weights are developed. If patient data collection is deficient in their hospital, they will appreciate how this affects cost weights. Access to reliable patient costing data allows benchmarking, which is essential if clinicians are to be effective managers and control costs, while maintaining and improving quality.
Victorian cost weights
	Cost weights in Victoria have been developed from hospitals with computer based costing systems. There were five hospitals in the initial study and 15 in the 1996-97 study. Having individual patient data coded according to ICD-9 has made it possible to identify unexpected variations and outliers, and to go back to the source hospital for data verification, especially for low volume DRGs. Knowledge of the adequacy of each hospital's costing system has been important in achieving valid cost weights. Variation in the cost weights from year to year has been partly due to the changes in the national classification system (AN-DRGs), to the impact of steadily improving hospital costing systems (by introducing new and improved feeder systems), to enhanced data collection, and to clinical variability within small volume DRGs from one year to the next. There have also been actual changes in reported costs because of gradually reducing length of stay, increases in day surgery and changes in clinical management practice (eg, use of different drugs, operative procedures or prostheses). An annual update of cost weights is essential, if hospitals are to be confident about the cost weights underpinning casemix payment formulas. Once systems are in place, this can be achieved at reasonable cost.
A simplified patient costing system
	Many hospitals do not have patient costing systems, because they cannot afford such a system or because their size does not justify the investment. A method has been developed that emulates the results of more expensive and sophisticated systems, but at a fraction of the cost. It is a database management system that draws together clinical and cost data at the patient level. This enables "bottom up"-based costing of patients which can then be summed to give overall costs and hence cost weights. Although this method does not yield the sophisticated data of the larger systems, clinicians often find themselves overwhelmed by data produced by larger systems, when in fact they would prefer fewer but more strategic and relevant data.
Conclusion
	Cost modelling has inherent limitations, and the alternative -- detailed patient costing -- requires efficient hospital information technology systems. However, both use the same cost disaggregation systems for overheads and medical costs and, and in hospitals without nurse dependency systems, also for nursing costs. The suggested simplified approach to patient costs may be attractive for smaller hospitals, as it avoids considerable capital expenditure. A better classification system and funding formulas incorporating reliable cost weights derived from "bottom up" patient costing systems should overcome many of the deficiencies identified by clinicians in the current casemix payments systems. It should better describe patients, particularly in teaching and specialist hospitals, and also smaller hospitals, each of which has a unique AN-DRG patient mix.
References
	Fetter RB, Thompson JD, Mills RE. A system for cost and reimbursement control in hospitals. Yale J Biol Med 1976; 49: 123-136.
Authors' details
	Faculty of Health Sciences, La Trobe University, Melbourne, VIC. Peter D Phelan, MD, FRACP, MRACMA, Adjunct Professor. Health Solutions Pty Ltd, Melbourne, VIC. Richard Tate, BA(SocSci), GradDipAccount, Group Executive. Healthwise Consulting, Mebourne, VIC. Fiona Webster, BSc, MBA, GradDipApplPsych, AFCHSE, Health Consultant. Acute Health Division, Department of Human Services, Melbourne, VIC. Ric P Marshall, BA, DipPsych, GMQ, Manager, Information and Performance Evaluation. Reprints will not be available from the authors. Correspondence: Professor P D Phelan, Victorian Medical Postgraduate Foundation, PO Box 27, Parkville, VIC 3052. E-mail: vmpfATvicnet.net.au

	Go to the next article . . . Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company. Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au>". <URL: http://www.mja.com.au/> © 1998 Medical Journal of Australia.