1. Jornadas de economia de la
salud
June 2010
Valencia
Measuring health
services (in the
national accounts)
Paul Schreyer, OECD
2. Background
• Much effort spent on measuring the value of
GDP at current prices
• Even more important: volumes.
• Difficult when
– Products are complex with changing quality
– There are no economically significant prices
3. Background
• Health services are a prime example:
– Rapid technological change
– Provision often by non-market suppliers
• So how are health services measured in the
NA?
• And should this be done differently?
4. Background
• Recent work:
– Eurostat (2001) & EU Regulation
– Atkinson report (2005) and work by
UKCMGA
– United States: Triplett and Bosworth 2004,
Abraham and Mackie 2006
– OECD:
• Handbook 2010
• Data development: health PPPs
– Stiglitz-Sen-Fitoussi Commission (September
2009)
5. This presentation
• A few concepts
• Capturing health services and quality
change
• Conclusions
6. Terminology
• Inputs
• Outputs
– Processes without explicit quality adjustment
– Processes with explicit quality adjustment
• Outcomes
– Direct outcomes
– Indirect outcomes
• Best explained by way of a graph
7. Inputs Outputs Outcomes
Process without Process with
Direct Indirect
explicit quality explicit quality
outcome outcome
adjustment adjustment
Example Example
education: education: quality- Knowledge
number of adjusted number and skills as
pupils/pupil hours of pupils/pupil measured by Future real
by level of hours by level of scores earnings,
education education growth rate
Labour, capital,
of GDP,
intermediate well-
inputs Example health: rounded
Example health:
quality-adjusted citizens
number of
number of Health status etc.
complete of population
complete
treatments by
treatments by type
type of disease
of disease
Inhereted skills, socio-economic
Environmental
factors
background, etc.
Hygene, lifestyle, infrastructure
etc.
8. Terminology
→National accounts should strive at
measuring outputs
→in the form of quality-adjusted processes
→or as the marginal contribution to outcomes
→Outcomes (as used here) is indicative of
results of the health system as a whole
→Although outcome ≠ output, they are not
independent
9. Non-market production
• Nothing special in terms of the basic
service provided, but mode of provision is
different
• In particular: prices not economically
significant
• Traditional way of dealing with this:
– Value of output = value of inputs (sum of
costs)
– Volume of output = volume of inputs
– Problem: defines away productivity change
10. From inputs to outputs
• Efforts to capture volume change of
outputs, not of inputs
• What exactly is output of health industry?
• Target definition:
• Health service = complete treatment
of a disease or medical service to
prevent a disease
• But what exactly is a ‘complete treatment’?
• And how to deal with quality change?
11. • The long (& still incomplete) answer is
here:
OECD (2010); Towards Measuring the
Volume Output of Education and Health
Services: A Handbook
• The short answer in the slides to follow.
12. Completed treatment
• Pathway that an individual takes through
heterogeneous institutions in the health
industry in order to receive full and final
treatment for a disease or condition.
• Not easily applicable, conceptually and
empirically:
– E.g., chronic diseases, data limitations
• After reality check: treatment of one
episode of disease by a particular
institution
• Main point: towards disease-based
estimates
13. Measuring treatments
• Required to construct disease-based
output measures:
– number of treatments by type of disease or
– costs per treatment by type of disease
• Administrative sources (e.g. DRGs) have
significantly facilitated access to such
information (Germany, Denmark,
Australia, France,…)
• Volume indices can then be constructed:
• Weighted average of # of treatments
• Weights: cost share of type of disease
14. Denmark: Input price index and output price index
for general hospital services
based on ~ 800 DRGs
120.0
115.0
110.0
105.0 Input based
Output based
100.0
95.0
90.0
2000 2001 2002 2003 2004 2005
15. Quality change
• Via stratification (implicit quality
adjustment)
• Explicit quality adjustment
• Capturing quality by measuring
marginal contribution to outcomes
16. Capturing quality change via
stratification (1)
• Basic idea: matching products: services are
stratified such that only similar services are
compared
• Criterion for classification: similarity in effects
on patients, and not similarity in how services
are produced
• Example: different treatments for the same
disease (in a particular type of establishment)
should be in one stratum, and not similar
medical procedures for different diseases
• Implies: most detailed stratification = not
necessarily best choice with cost weights and
imperfect markets
17. Capturing quality change via
stratification (2)
• 2 treatments, traditional and laser surgery
Same effect on outcome
Laser is cheaper
Laser surgery diffuses progressively
Traditional surgery Laser surgery
Period 0 Period 1 Period 2 Period 0 Period 1 Period 2
Unit cost 100 100 100 - 90 90
Number of interventions 50 40 5 0 10 45
Total cost 5000 4000 500 0 900 4050
Laspeyres volume index period 1 to period 2:
[sT(5/40)+sL(45/10)]-1=-7.1%
where sT=82% and sL=18% are the period 1 cost shares
18. Capturing quality change via
stratification (3)
• Measured output declines! →Counter-intuitive
• Problem: 2 medical procedures have been
treated as 2 different services
• Note also implicit assumption:
– consumer valuation of the two ‘products’ is
captured by the relative unit costs,
– so if laser surgery is cheaper than traditional
surgery, this method implicitly quality-adjusts
downward the quantity of laser surgery when
it is combined with traditional surgery.
19. Capturing quality change via
stratification (4)
• In a perfect market, the price of the
traditional treatment would
instantaneously adjust downward and/or
traditional treatment would disappear
• In practice and in a non-market context,
this does not happen
→2 treatments should be treated as the
same product
→no cost weighting, volume change = zero
20. Capturing quality change via
stratification (5)
• A more sophisticated method would be to keep
both treatments in the same stratum but
explicitly adjust one of the treatments
• For example put a coefficient of 0.8 on the
traditional treatment
• Such adjustments would have to rely on medical
effectiveness studies
• There is some way to go before practical
implementation
• For more on this, see Triplett (2001): What’s
different about health? Human repair and car
repair in national accounts and in national
health accounts
21. Capturing quality change via
stratification (6)
• Note: even if we do not explicitly quality-adjust products,
the decision how to group them cannot be made without
some reference to effects on outcome
• Otherwise, no statement can be made about
substitutability of services and how they should be
classified
• Nearly everything that has been said about quality
adjustment via stratification applies also to market
production
• Only difference: price statisticians or national
accountants who have to deal with it
22. Capturing quality change via
explicit adjustment
• If stratification is too granular, explicit
quality adjustment may be needed
• Basic problems:
– Quality is multidimensional: how to aggregate
across dimensions? (waiting times, risk of
suboptimal treatment...)
– How to aggregate change in quality variable(s)
with change in quantity (process) variables?
23. Capturing quality change via
measuring marginal contribution to
outcome
• Example: QUALYs, DALYs
• Problems:
– Making sure factors other than health care are
controlled for;
– Timing – lag between service and outcome
– Putting monetary value on human lives
• Very useful avenue of research but unlikely to be
implemented in official statistics soon
24. Conclusions (1)
• Output and outcome are different and should
not be confused
• National accounts and productivity studies of
establishments need measures of output, not of
outcome
• But output and outcome are not independent
• In the presence of quality change, all existing
methods require some implicit or explicit
information or reasoning about outcome.
• Output or medical services should be based on
treatment of disease unit of production
25. Conclusions (2)
• Problems of quality adjustment arise
whether services are provided by market
or non-market producers.
• A pragmatic approach will be called for to
proceed with services measurement
– no reason to approach every type of service
with the same method for quality adjustment
– methodologies should be robust and
transparent
26. Conclusions (3)
• Measuring output for complex services is
difficult
• But conclusion should not be that it is simply too
difficult to do anything.
• Visible progress in the area of health due to
availability of administrative data
• It may take a while before consensual and
internationally comparable methods are agreed
upon but active research and data development
is vital to achieve this objective.
