Enhancing Output Measures of the Health Care Industry
By Qinghuan Luo, Economic Research Section
The health care industry is characterised by rapid innovation and advances in medical technologies, such as the development of new drugs, or new procedures, that are less invasive and achieve better outcomes for patients. These changes pose major challenges for national accountants and health economists in measuring price change or volume change in health care output. Measurement of health care output is further complicated by a general lack of market transaction prices for the non-market sector of health care.
The Australian Bureau of Statistics (ABS) currently uses the direct quantity index method for estimating volume change in health care output, in line with international practice in national accounting. The current method was developed in early 2000s, where a quantity index is constructed using activity-based quantities (such as the number of hospital separations), weighted by costs.
There has been some progress in this field internationally, notably the development of output measures based on disease treatment, and adjustment to output for quality of health care, to better address the challenge of accounting for quality change. The US Bureau of Economic Analysis (BEA) and the Bureau of Labour Statistics (BLS) have recently developed experimental disease based price indexes. BEA has constructed a disease based Health Care Satellite Account. The UK’s Office for National Statistics (ONS) has developed an explicit quality adjustment approach, and applied it for productivity estimates.
The recent progress in this field presents opportunities for enhancements to the ABS’s current approach for measuring health care output. The ABS has reviewed the international experiences, and has examined various options for improving output measures for Australia.
This paper describes the options for improving health care output measures for Australia, and outlines the proposed developments. In the medium term, the ABS will focus on developing a partial disease based approach, where an output volume measure is constructed based on disease treatment by type of providers (such as hospitals and primary care providers) classified by the Australian and New Zealand Standard Industrial Classification (ANZSIC) Class or Subdivision. The improved output volume measure will be used to construct improved productivity estimates of the health care industry.
The health care industry commands an increasingly significant share of the Australian economy, and is an important source of economic growth. Understanding the key drivers for changes in price and volume measures of health care output is of interest for policy makers and researchers in the context of economic growth.
A majority of health care services are funded (either directly or indirectly) by the Commonwealth and state/territory governments, and spending on these services comprises a large component of overall government expenditure. As such, reliable statistics for health care output and productivity are important for government policy makers to gain insights into growth in health expenditure (e.g. whether spending growth is due to price growth or volume growth), and industry wide productivity performance. Note that apart from productivity there are other performance indicators such as efficiency, which are of particular relevance for health policy makers but are beyond the scope of this paper. The ABS seeks to address the need for statistics for health care, by focusing on enhancing output estimates for health care industry.
The health care industry considered here is classified under Subdivisions 84 Hospitals and 85 Medical and Other Health Care Services of the Australian and New Zealand Standard Industrial Classification (ANZSIC) Division Q Health Care and Social Assistance (see Appendix 1 for detail of the ANZSIC Classes in scope).
This paper provides an overview of the general methods for estimating volume measures for health care services and the key challenges, and discusses different options for enhancing measurement of health care output for Australia. The paper outlines the ABS’s proposed developments in enhancing volume estimates for health care output.
2. MEASUREMENT OF HEALTH CARE OUTPUT
GENERAL METHODS FOR VOLUME ESTIMATES
Specific methods for estimating health care volume output for national accounts are largely determined by available data. As health care comprises both market and non-market sectors, both deflation and direct volume methods have been used in national accounting (footnote 1).
The deflation method
The output of many industries in the national accounts can be measured on the basis that output and prices can be determined from transactions between producers and users (or consumers). An output price index can then be constructed by pricing identical products from period to period. The volume measure of output can be estimated by deflation of the current price value using the price index.
In principle, an output price index for the market sector of health care can be constructed using this approach.
For the non-market sector, due to the lack of a transaction price, it is not possible to construct price indexes in the usual Producer Price Index framework. One option is to construct a “pseudo output price index”, which can be based on an input price index, adjusted for growth in productivity, or output price indexes of similar types of products that have prices.
Alternatively, a price index for the non-market sector can be constructed directly using unit cost of output as a proxy price, and a volume measure is then derived by deflation. A price index based on unit cost is sometimes referred to as the quasi-price index, or indirect price index, as it can be derived by dividing the total costs by the quantity index (Schreyer 2012). There is an important difference between the unit cost of output and the unit cost of input. The former encompasses an output element and can be used as pseudo output prices. By contrast, the latter is basically input prices, e.g. hourly wage of labour input to the treatment. Such differences have implications for measuring productivity for non-market production (Diewert 2017). Growth in the volume of output can be different from growth in the volume of input even though one assumes that total output value equals total input value.
The direct quantity index method
Volume change can be calculated by constructing a quantity index based on output quantity indicators, weighted by costs. This method is referred to as the direct quantity index method, and is recommended by the System of National Accounts 2008 (SNA 2008) for calculation of volume change for non-market output (SNA refers this method to as the “output volume method”).
The commonly used output indicators are activity-based quantities, such as the number of hospital separations and the number of visits to General Practitioners (GPs) or specialists. A quantity index can be constructed by stratification of activities at an appropriate level of detail, depending on data availability. For example, hospital services can be grouped based on a patient classification system such as Diagnosis Related Group (DRG) (footnote 2).
The direct quantity index method captures some aspects of quality changes, implicitly through both stratification of products (e.g. treatments) with similar characteristics and the regular updating of the weights. For example, an increase in the use of new or improved treatments that are initially more expensive than existing treatments may reflect partially in an increase in the number of such treatments and the weight of input cost. This would lead to an overall increase in the output volume.
One limitation of a cost weighted quantity index is that cost weights may not fully reflect patient valuation of the treatment in the context of treatment outcome. A quantity index may not capture substitution of a new, less expensive treatment for the traditional, more expensive treatment. This second limitation may be partially overcome by re-defining output by disease treatment (Groshen et al. 2017).
HEALTH CARE OUTPUT IN THE ABS NATIONAL ACCOUNTS
The ABS compiles health care outputs by industry and by products, as for other industries, in the supply-use framework of the National Accounts.
While the output of the market sectors is usually estimated from the supply side, using output data such as sales revenue, estimates of current price value of health output for both market and non-market sectors are currently derived from the demand side using a variety of data sources including survey data and administrative data. The current price value of output is estimated as the sum of intermediate consumption of health care services and final demand (final expenditure and exports less imports). The ABS has ongoing work to develop supply-side estimates for current price value of health care.
As part of the compilation process for volume measures of health care services, the ABS compiles health care quantity indexes based on the output volume method for non-market producers. The method was developed in early 2000s, based on cost weighted quantities of health care activity, which is consistent with the recommendation in SNA 2008. A hospital quantity index is calculated as the cost weighted number of admitted patient separations, and quantity indexes for non-hospital services are calculated by types of services. The number of hospital separations by procedure type, and average separation costs, are sourced from the Australian Institute of Health and Welfare (AIHW) hospital publication. The number of non-hospital services provided and costs are sourced from Medicare, the Private Health Insurance Administration Council and the Productivity Commission (PC) Report on Government Services. The current approach is not explicitly adjusted for quality change. (For further information, see Chapter 9, Australian System of National Accounts: Concepts, Sources and Methods, ABS cat. no. 5216.0, 2015.)
The ABS does not currently produce health care productivity statistics, and intends to construct productivity measures as part of development of improved output volume measures.
Concept of health care output
It is generally accepted that the unit of health care output should be defined as complete treatments (National Research Council 2010, OECD, Eurostat and World Health Organisation 2017). For example, a complete treatment of a hip fracture patient encompasses services provided by different providers. In this case, the patient is initially treated at an Emergency Department (ED), subsequently undergoes an operation in an Orthopaedic Ward of the same hospital or another hospital as an admitted patient, and undergoes rehabilitation at an allied health provider. The cost for the complete treatment includes costs for ED and Admitted Patient Care (APC) in the hospital, and allied health care.
A practical challenge of implementing this ideal concept of health care output is the requirement for data with linkages across different types of providers. Such fully linked data is not available in Australia. A second challenge is that integration of disease based measures into national accounts is not straightforward, as outputs in national accounts are estimated by types of products by types of providers (as opposed to disease based measures calculated across different types of providers).
National Statistical Offices (NSOs) usually adopt a pragmatic approach of measuring health care output as the number of episodes of treatments or events of patient care by types of providers (Schreyer 2010). Outputs associated with ED events and APC episodes in hospitals, and physiotherapy treatment, are measured separately and aggregated by types of providers, usually classified by an industry classification.
The US Bureau of Economic Analysis (BEA) and the Bureau of Labour Statistics (BLS) have recently developed experimental disease based price indexes. BEA has constructed a disease based Health Care Satellite Account where output is measured based on disease (Dunn, Rittmueller and Whitmire 2015). Their experience showed that full disease based measures can be constructed where patient level data with linkages across different providers is available. To produce statistics in practice, a pragmatic strategy is needed in dealing with complications such as treatment of multiple medical conditions (see Appendix 2).
Australian health care comprises both market and non-market sectors, with the latter being mainly funded by government. Measurement of health care output for the non-market sector is further complicated by a general lack of market prices. As non-market health care services are often provided to consumers at zero prices or at prices below input costs, output cannot be valued directly at market transaction prices.
In the absence of market prices, it is generally accepted that the non-market output is valued based on input costs or expenditure (SNA 2008). Thus, the current price value of output is estimated as the total cost for supply of the services or the total expenditure by households and government. This assumption underpins the usual approach for deriving an output volume measure, i.e. use of cost or expenditure weights in constructing a quantity index (or use of unit costs in constructing a price index which is then used for deriving a volume measure).
It is worth mentioning that even for the private sector of health care, market prices may not be readily available, and costs for providing these services are used for output measurement. For example, costs and quantities by DRG for private hospitals are routinely collected by the health authorities in Australia, and the data can be used for constructing a quantity index (or a price index based on unit cost) in the same way as for public hospitals.
The health care industry has experienced rapid advances in medical technologies and innovation. New treatments involving use of new drugs and advanced procedures frequently become available. These changes present major challenges for estimating price changes or volume changes in output for both market and nonmarket producers.
A quality change in health care may be reflected as substitution of a new, less expensive but more efficacious, treatment for the traditional, more expensive treatment. A quality change may also be reflected as a shift across different types of providers, say a new treatment that can be performed at an outpatient clinic and costs less than the traditional treatment, usually performed at a hospital as an inpatient.
These substitution effects can be largely captured by measuring output by disease treatment, such as the BEA disease based method (Dunn, Rittmueller, and Whitmire 2015). In some way, this approach encompasses implicit quality adjustment, as it partially accounts for new or improved treatments through capturing substitution in the resultant volume or price measure of output.
Although re-defining output by disease treatment can partially address the challenge of accounting for new or improved treatments, explicit quality adjustment is needed to fully capture changes in quality of health care, especially those reflected as a change in treatment outcome.
A general issue with explicit quality adjustment is to identify an appropriate set of characteristics that define quality, as economic theory does not specify which set of characteristics should be selected for defining quality of health care. Selection of appropriate quality indicators requires medical knowledge and information on how treatments are valued by consumers.
It is generally accepted that quality adjustment may require taking into account health care outcomes in some way (e.g. Triplett 2001, Atkinson 2005, National Research Council 2010, Schreyer 2012, Diewert 2017). As summed up in Schreyer (2012), “outputs should reflect the results of production and these cannot normally be captured by outcome, the state of health [or education] of the population, However, […] outcome information is required when it comes to quality adjustment of output measures”.
The key challenge of an outcome-based quality adjustment is to determine the marginal contribution of output to health outcome. This requires netting out any effects that are not due to medical treatments, such as improvement in health due to non-medical factors related to improved diet or quitting smoking. See Appendix 3 for more discussion of quality adjustment.
3. ENHANCING HEALTH CARE OUTPUT MEASURES
There are a number of options for enhancing output measurement of the health care industry by going beyond the current ABS method of using input cost weighted quantity of activity. These include:
1. Partial disease based output measures by industry
2. Explicit quality adjustment to the input cost weighted quantity of activities (the ONS approach)
3. Full disease based output measures (the BEA approach)
4. Explicit quality adjustment to disease based output measures (ideal case)
Each option would allow the construction of improved productivity measures.
Option 1 is a scaled down version of the BEA disease based approach (Appendix 1), with disaggregation to disease treatment limited to within ANZSIC Class or Subdivision. This approach is favoured by the ABS, as it enables implicit quality adjustment by capturing substitution across different types of services and different providers classified to different ANZSIC classes. It is more feasible than the full disease based approach. For example, a disease based measure can be constructed for hospital services as linked patient data across different types of hospital services (such as ED, non-admitted patient care and APC) is potentially available.
Output measures with disease based disaggregation confined within an industry class or subdivision can be easily integrated into the existing supply-use framework in the National Accounts by including a dimension of disease treatment under each ANZSIC Subdivision 84 and 85.
Some countries, such as the Netherlands, have already adopted this approach to various degrees. Statistics Canada has developed an experimental output measure for hospitals, taking into account a compositional shift from inpatient care toward outpatient care in hospitals (Gu and Morin 2014).
Option 2 is the ONS approach (Appendix 2). Quality adjustment based on outcome indicators is applied in construction of activity based quantity indexes where, as in the conventional approach, input costs are used as weights. One advantage of this option is to make use of the existing health care quantity indexes, with the main work being the development of outcome based quality indicators.
A weakness of this option is that the underlining quantity index is not disease based. The index may not capture substitution effects, such as a shift between different types of providers. It is noted that the ONS currently implemented quality adjusted output measures in producing productivity estimates for health care, not in their national accounts.
Option 3 is similar to the BEA disease based approach. It captures substitution across different providers or different types of services, but without explicit adjustment for treatment outcomes. This option requires completely linked patient level data in the form of time series. Although there is a growing demand for data linkage across different types of providers in the health sector, and such linked data will become gradually available in the future, this option is currently not practical in Australia. Nonetheless, this approach can be considered in the future when such data become readily available. It is worth noting that there are studies of expenditure by disease for Australia (e.g. AIHW 2010).
It is not easy to integrate full disease based measures into the existing national accounts, where outputs are usually compiled by industry or types of product.
Option 4 is ideal, as it captures substitution effects and includes quality adjustment. As in Option 3, it is not easy to incorporate the full disease based measures into the existing national accounts. It demands fully linked patient level data which is not available. Thus, this path is currently not practical.
In implementation, a health care satellite account is highly desirable. It can bridge a gap between the health care industry in the National Accounts and AIHW health expenditure statistics, and facilitate compilation of health care measures in the National Accounts (SNA 2008). The satellite account can be used as a “test ground” for implementing extension of concepts, such as fully disease based measures and outcome based quality adjustment.
The availability of relevant data is the key to the development for each of the options being feasible. While administrative datasets are available in the health care industry, these datasets are disparate and difficult to gain access to and to link to other data sets. In Australia, public health care services are funded by the Commonwealth and state/territory governments. While government expenditure data at an aggregate level is available, access to lower level data on costs and activity (such as those collected from hospitals) requires negotiation with the designated data custodians.
Compilation of quantity indexes (or price indexes based on unit costs) requires a fine level of quantity and cost data. Health care administrative data vary in detail by different sectors, among which hospital data is the most comprehensive and detailed. Hospitals routinely collect patient data which contain patient level information such as personal details, diagnosis and procedure or treatment, including costs for individual patients who receive admitted acute care. In Australia, two important data collections are the AIHW National Hospital Morbidity Database (NHMD), which comprises episode level records of admitted patient care, and the Independent Hospital Pricing Authority (IHPA) National Hospital Cost Data Collection (NHCDC), which contains patient level cost data.
By comparison, data on non-hospital services (e.g. general practice, specialist medical services) are usually collected by providers such as Medicare and private health insurers. The available data contain less detailed information on diagnosis or treatment than hospital data.
The ABS currently uses administrative data in compiling health care quantity indexes. However, improvement beyond the current input-cost weighted approach requires greatly expanded use of the available administrative data, as well as exploration of alternative data sources. For example, development of disease based measures requires tackling the challenge of linking patient data across different types of providers. Quality adjustment requires exploring the capability of NHMD and NHCDC for developing quality indicator time series.
4. ABS’S PROPOSED DEVELOPMENTS
Disease based output measure
In the medium term, the ABS will focus on developing partial disease based output measures, i.e. Option 1, where output price and volume measures are constructed based on disease treatment by industry (e.g. hospitals versus primary care providers). This option is the preferred development path, as it is disease based (limited to types of providers) and encompasses implicit quality adjustment (see Section 3). Table 1 summarises potential data for developing disease based output measures. Note that listed sources are not exhaustive.
The ABS will initially develop disease based output measures for hospitals classified to ANZSIC Subdivision 84. The key hospital datasets are NHMD and NHCDC (Table 1). Methods for linking different types of hospital services, e.g. inpatient care and outpatient care, will be investigated.
The ABS will work closely with data custodians (AIHW and IHPA) on how to maximise value of NHMD and NHCDC data for the purpose of statistics production.
The ABS will determine appropriate levels of disaggregation of treatments in the construction of quantity and price indexes. An index can be constructed from the patient level versus a more aggregated level, say AR-DRG. Both quantity index and price index (based on unit costs) for hospital services will be constructed.
For non-hospital services (in scope of ANZSIC Subdivision 85), the ABS has identified MBS and Private Health Insurance data as two key datasets. The main challenge is to allocate these services to disease treatments, as the data on some types of services in MBS and the published Private Health Insurance data do not contain details on disease treatments.
To address this challenge, the ABS will explore options for the allocation of the services to disease treatment, including sourcing additional data. For example, GP services may be disaggregated to disease treatment using the Bettering the Evaluation and Care of Health (BEACH) survey (Britt et al. 2016).
The improved output volume measures will be used to derive productivity estimates for the health care industry.
Initially, the ABS will undertake work on developing disease based output measures for hospitals and expects to produce experimental quantity indexes in late 2018.
Quality adjusted output measure
The ABS will continue to consult with health experts, such as health economists, on the best way forward for developing a quality adjustment approach that is both conceptually sound and practical for statistics production.
The ONS experience showed that construction of quality adjusted output measures is very challenging, especially for Australia, mainly due to a lack of data on measurement of health effects attributable to medical treatments. In Australia’s case, the key hospital datasets (Table 1) may be used for developing quality indicators for hospitals. However, various indicators need to be incorporated into the Quality Adjusted Life Year (QALY) or Disability Adjusted Life Year (DALY) measures (Appendix 3). There is currently no internationally agreed framework for adjustment of output for treatment outcomes.
The ABS considers that the best way forward is to work closely with both academics in this field and key data custodians, with options to engage health economists to assist in developing quality indicators that can be used for construction of quality adjusted quantity indexes.
Table 1: Major data sources for constructing disease based output measures
|Major data sources||Details||Use for quantity indexes||Use for quality indicators|
|AIHW National Hospital Morbidity Database (NHMD)||Episode level records of admitted patient care, based on Admitted Patient Care NMDS and Admitted Subacute and Non-acute Hospital NBEDS.||The number of separations can be used for compiling a quantity index for hospitals, using cost weights sourced from NHCDC.||E.g. In-hospital Mortality Rates; Primary Care Indicators - avoidable hospital separations; Patient Safety Indicators (AIHW 2016).|
|IHPA National Hospital Cost Data Collection (NHCDC)||Patient level cost data contains five streams: Admitted Acute, Emergency Department (ED), Non-admitted, Subacute, and Other.||Costs and number of separations can be used as weights for compiling a quantity index for hospitals.||E.g. Hospital Acquired Complication (HAC).|
|National Non-admitted Patient Emergency Department Care Database (NNAPEDCD)||Episode level non-admitted patient ED compiled from Non-admitted Patient Emergency Department Care NMDS, and Non-admitted Patient Emergency Department Care NBEDS.||Provide quantity for ED care.|
|Medicare Benefit Scheme data||Medicare Benefit Schedule (MBS) claim data.||Additional data is needed for some types of services (e.g. GPs) to be allocated to disease treatment.|
|Private Health Insurance data||Expenditure and quantity by broad types of services.||Additional data is needed for allocating services to disease treatment.|
A number of key challenges exist in the measurement of health care output. The most challenging aspect of measuring output is to account for quality changes, as the health care industry continues to experience rapid change in technology. This challenge is further compounded by the lack of market transaction prices for valuing output with quality adjustment for the non-market sector. Access to, and maximising the use of health administrative data for producing statistics presents a further challenge.
The ABS has commenced a project on developing enhanced output measures for health care by addressing these challenges. Initially, the ABS will focus on developing partial disease based output price and volume measures, where disaggregation by disease treatment is limited to an ANZSIC Class or Subdivision. The improved output volume measures can be implemented in the National Accounts and will be used for constructing productivity measures for the health care industry. The ABS will consult with health economists on the best way forward for developing quality adjustment, with a long term goal of including quality adjustment in the output measures.
Atkinson, T. 2005. Atkinson Review: Final report. Measurement of Government Output and Productivity for the National Accounts. Hampshire, England: Palgrave-Macmillan.
Australian Institute of Health and Welfare (AIHW). 2016. OECD Health-Care Quality Indicators for Australia 2015. Canberra: AIHW.
Australian Institute of Health and Welfare (AIHW). 2010. Health System Expenditure on Disease and Injury in Australia, 2004-05. Health and welfare expenditure series no. 36. Cat. no. HSE 87. Canberra: AIHW.
Bradley, R., Hunjan, J., and Rozental, L. 2015. Experimental Disease Based Price Indexes. https://www.bls.gov/pir/journal/rb03.pdf
Britt, H. et al. 2016. General Practice Activity in Australia 2015-16. General Practice Series no. 40. Sydney: Sydney University Press.
Dawson, D, Gravelle, H, O’Mahony, M. et al. 2005. Developing New Approaches to Measuring NHS Outputs and Activity. Centre for Health Economics University of York and the National Institute of Economic and Social Research. CHE Research Paper 6.
Diewert, W.E. 2017. Productivity Measurement in the Public Sector: Theory and Practice.
Dunn, A., Rittmueller, L., and Whitmire, B. 2015. Introducing the new BEA Health Satellite Account. Survey of Current Business 95 (1).
Gold, M.R., Stevenson, D., and Fryback, D.G. 2002. HALYs and QALY and DALY, Oh My: Similarities and Differences in Summary Measures of Population Health, Annual Review of Public Health, Vol. 23, 115-134.
Groshen, E.L., Moyer, B.C., Aizcorbe, A.M., Bradley, R., and Friedman, D.M. 2017. How Government Statistics Adjust for Potential Biases from Quality Change and New Goods in an Age of Digital Technologies: A View from the Trenches. Journal of Economic Perspectives, Vol. 31, 187-210.
Gu, W. and Morin, S. 2014. Experimental Measures of Output and Productivity in the Canadian Hospital Sector, 2002 to 2010. The Canadian Productivity Review. April 2014. Statistics Canada. (Catalogue no. 15-206-X)
Hall, A.E. 2017. Adjusting the Measurement of the Output of the Medical Sector for Quality: A Review of the Literature. Medical Care Research and Review, Vol. 74, 639-667.
National Research Council, 2010. Accounting for Health and Health Care: Approaches to Measuring the Sources and Costs of Their Improvement. Washington, DC; The National Academies Press.
OECD, Eurostat and World Health Organisation, 2017, A System of Health Accounts 2011, revised edition, OECD Publishing, Paris.
Office for National Statistics (ONS). 2012, Quality Adjustment of Public Service Health Output: Current Method. UK ONS Information Note, April.
Schreyer, P. 2010. Towards Measuring the Volume Output of Education and Health Services: A Handbook. OECD Statistics Working Papers, 2010/02, OECD Publishing, Paris.
Schreyer, P. 2012. Output, Outcome, and Quality Adjustment in Measuring Health and Education Services. Review of Income and Wealth, Vol. 58, 257-278.
Triplett, J. E. 2001. What’s Different about Health? Human Repair and Car Repair in National Accounts and National Health Accounts. in D.M. Cutler and E.R. Berndt (eds.), Medical Care Output and Productivity (National Bureau of Economic Research Studies in Income and Wealth Series, Volume 62, 15-94), Chicago, IL: University of Chicago Press.
APPENDIX 1: HEALTH CARE ACTIVITY UNDER ANZSIC 2006
Health care activity in scope is classified under the Australian and New Zealand Standard Industrial Classification (ANZSIC 2006) to Subdivision 84 Hospitals and 85 Medical and Other Health Care Services.
Table A1: Health care component of ANZSIC Division Q - Health Care and Social Assistance.
|8401 Hospitals (except Psychiatric Hospitals)|
|8402 Psychiatric Hospitals|
|85 Medical and Other Health Care Services:|
|8511 General Practice Medical Services|
|8512 Specialist Medical Services|
|8520 Pathology and Diagnostic Imaging Services|
|8531 Dental Services|
|8532 Optometry and Optical Dispensing|
|8533 Physiotherapy Services|
|8534 Chiropractic and Osteopathic Services|
|8539 Other Allied Health Services|
|8591 Ambulance Services|
|8599 Other Health Care Services n.e.c.|
APPENDIX 2: INTERNATIONAL EXPERIENCE
There are extensive studies on measurement of health output and quality adjustment in the literature (Triplett 2001, Dawson et al. 2005, Schreyer 2012, Diewert 2017, Hall 2017). The US Bureau of Economic Analysis (BEA) full disease based approach and the Office for National Statistics (ONS) quality adjustment approach represent two very different paths for improving upon the existing approaches. Note that BLS also constructed experimental disease based price indexes (Bradley et al. 2015). The BEA and the ONS approaches are discussed here.
The BEA disease based output measures
The key feature of the BEA approach is to measure output by disease treatment (Dunn, Rittmueller and Whitmire 2015). This is done by tracking a patient journey and measuring the output as spending per patient per medical condition in a calendar year. A patient may receive a range of services from different providers and the total cost is the aggregate of expenditure for all services the patient received. The main data sources are the Medical Expenditure Panel Survey (MEPS) and medical claims data, which are patient centred.
In constructing the Satellite Health Care Account, BEA disaggregated expenditure by disease treatment. Price indexes were constructed based on disease treatment using the patient centred data (mainly MEPS and medical claims data), where price of treating condition is the average expenditure per patient for the condition. BEA derived diseased based volume measures by deflation.
The BEA approach is conceptually closer to the ideal definition of health care output, as it captures mostly complete treatments. However, their approach includes incomplete treatments that are administered during a period across two consecutive years. The BEA disease based output price indexes were not explicitly adjusted for treatment outcomes. The BEA has ongoing research on quality adjustment in health (e.g. Hall 2017).
The BEA Health Care Satellite Account is not integrated with the National Income and Product Accounts (NIPA). One complication of integration of disease based measures in the national accounts arises from the fact that disaggregation of expenditure across disease differs from disaggregation by types of goods and services. While the spending total should remain the same, different disaggregation structure can lead to differences in an aggregated price index.
The ONS outcome based quality adjustment
The ONS adjusts health care outputs using five quality measures developed by Centre for Health Economics, University of York (Dawson et al. 2005, ONS 2012). These include: health gain; short term survival rate; waiting times; patient experience from the National Patient Survey; and primary care outcomes.
Health gain is measured as a change in QALY after medical intervention, where QALY is discounted to take into account time preference. Short term survival is incorporated into health gain over the remaining life of those treated. The effect of waiting times is treated as loss in health gain, as a result of delaying treatment. Construction of QALY requires time series on health effects which may not be readily available. A “snapshot” of QALY is used instead of an actual time series.
The three measures of health gain, short term survival and waiting times as a combined measure are applied to Health Resource Group (HRG) activity, while patient experience is applied to most sectors. (HRG is a patient classification system similar to AR-DRG used in Australia.) Adjustment for primary care outcome is applied only to GP consultations. The growth factors of Patient Experience and Primary Care Outcome for GPs are combined multiplicatively.
As the ONS approach is largely outcome based, it has the usual limitations that other factors not related to medical intervention may contribute to a health outcome. Ideally such nonmedical effects should be netted out. For example, in controlling for nonmedical factors in primary care outcome, the ONS assumes that one third of growth in the proportion of relevant patients whose cholesterol levels are within the desirable clinical target is due to improvement in primary care and the rest attributed to factors not related to primary care (ONS 2012).
The ONS implemented quality adjusted health output measures in production of productivity statistics, but not in the National Accounts.
The ONS experience showed that constructing QALY is a big challenge due to lack of measurement data on health effects before and after treatment.
The ONS also showed collaboration between NSOs, health economists and health data custodians plays a key role in developing improved health output measures. For example, the ONS developed quality adjusted output measures in partnership with York University and Department Health.
APPENDIX 3: QUALITY ADJUSTMENT OF HEALTH CARE OUTPUT
This Appendix explains the basic concept of quality adjustment, and why an outcome-based approach is the more feasible approach for quality adjustment in health care output.
There are extensive discussions in the literature on quality adjustment, mainly in the context of price indexes. Quality adjustment makes a change in the price so that a change in quality is correctly reflected in the volume measures. In the context of quantity indexes, quality adjustment applies an adjustment factor to quantity to reflect a change in quality. The following discussion is applicable in both contexts of price index and quantity index.
There are two broad categories of quality adjustment in health care: 1) process-based, where quality change is measured using indicators directly related to production processes of health care services such as details of a specific treatment or procedure performed; and 2) outcome-based, where quality change is measured using observed change in health outcomes (Hall 2017).
In discussion of quality adjustment in health services, it is useful to distinguish between the concepts of output and outcome in the national accounts. The output is what is produced, i.e. the result of production. The outcome is a state valued by consumers, which is outside the production boundary in the national accounts (Schreyer 2010).
The process-based approaches are in line with the output concept of the national accounts. While being conceptually preferred in the national accounts, a purely process-based approach is difficult to implement in practice due to lack of detailed data on individual treatments.
The main basis for the relevance of health outcome in quality adjustment is the marginal contribution to direct health outcome by medical intervention, where health outcome or state of health can be a function of treatments and other non-medical factors. This means that in theory one may track change in output by tracking change in health outcome, with other non-medical factors being held constant (Triplet 2001, Schreyer 2010). SNA 2008 suggests that “one way of identifying explicit quality adjustment factors is by reviewing the effects that the service has on measures of outcome” (paragraph 15.122, SNA 2008).
A range of health care outcome indicators can be relevant for quality adjustment, such as Quality Adjusted Life Year (QALY), Disability Adjusted Life Year (DALY), mortality, or survival rate. Among them, QALY and DALY are two widely used metrics for measuring health care outcome in the health economics literature. The QALY measure combines life quality and years of life, where years of life are weighted by life quality on a scale from 1 (perfect health) to 0 (death) (Gold, Stevenson and Fryback, 2002). One QALY corresponds to one year in perfect health. A gain in health as the result of medical intervention is measured as a gap between QALY with intervention and QALY without intervention.
DALY is a measure that combines years of life lost due to premature mortality and years of life lost due to time lived in health states less than ideal health (Gold, Stevenson and Fryback, 2002). The metric is increasingly used for measuring burden of disease. For example, DALY is used in the Global Burden of Diseases Study (GBD) and the Australian Burden of Disease Study (footnote 3).
Use of QALY and DALY in quality adjustment in health is relatively recent. The measures are considered useful indicators for quality changes (National Research Council, 2010). One of the advantages of using these measures is that they incorporate a measure of health gain in the future. This is important, as health benefit from treatment can extend well into the future. These measures are more robust in capturing new treatments than are direct outcomes such as mortality rate or survival rate.
However, construction of these measures can be challenging, due to data requirements for measuring changes in health effects attributable to treatments. In practice, choice of quality indicators could well be limited by availability of data as a time series.
1. Other methods, such as “input method” using changes in input volume as a proxy for output for non-market producers, are not discussed here.
2. DRG groups patients in a clinically meaningful way by relating the types of patients in a hospital according to their diagnoses, surgical procedures, and other routinely collected information such as ages and length of stay in the hospital. The Australian version of DRG is Australian Refined DRG (AR-DRG).
3. Australian Burden of Disease Study: Impact and causes of illness and death in Australia 2011. Australian Burden of Disease Study series no. 3. BOD 4, Australian Institute of Health and Welfare 2016.