Assessing productivity in Australian health services delivery: Some experimental estimates

1
Assessing productivity in Australian health
services deliverySome experimental estimates

• Owen Gabbitas and Christopher Jeffs
• Productivity Commission
• 17 December 2007

PRELIMINARY WORKING PAPER NOT FOR QUOTATION
WITHOUT PRIOR CLEARANCE FROM THE CORRESPONDING
AUTHOR,OWEN GABBITAS (ogabbitas_at_pc.gov.au)
2
Outline of presentation

• Setting the scene
• Conceptual framework for the delivery of health
  services
• What is productivity?
• Quality is an important aspect of healthcare
• State variation in average public hospital costs
• Stochastic frontier analysis of state territory
  public hospital systems
• Summary

3
Setting the scene

• The Commission gave an undertaking in Australias
  Health Workforce to pursue further work in the
  area of productivity measurement in health
  services delivery
• Our paper explores the availability and
  suitability of Australian health data for use in
  productivity analysis
• It looks at productivity at 3 levels in the
  health system
• health and community services (the health system
  in aggregate)
• public hospitals (the health service provider
  level)
• diagnostic categories related to hip replacement
  surgery(the procedural level)
• Focus today on public hospitals

4
Conceptual framework
5
What is productivity?

• Units of output per unit of input
• Concerned with physical units
• Does not take into account input or output prices
• Expressed in levels or, more commonly, growth
  rates
• Related to technical efficiency
• Extent to which inputs can be reduced while
  producing the same output (input-augmenting)
• Extent to which output can be increased from
  existing inputs (output-augmenting)
• Productivity focus is on measurement
• Policy focus is on efficiency and effectiveness

6
Quality is important

• Quality is multi-dimensional
• Quantity and quality of life (mortality
  morbidity)
• Quality may vary over time (inter-temporal
  nature) (eg survival rates)
• Indicators may also reflect other factors
  (attribution) (eg lifestyle)
• Choice of counterfactual?
• Before and after treatment
• What would otherwise have occurred
• Choice of appropriate quality measures to use?
• Composite measure based on indicators
• How to weight different metrics time periods?
• Overarching measures (eg life expectancy)?
• Can be incorporated into productivity analysis in
  various ways
• Through use of quality-adjusted output
• As a separate output in its own right
• Using the resulting health outcomes instead of
  outputs
• Seldom done in practice due to the absence of
  suitable summary measures

7
Considerable variation between treatments and
jurisdictions
8
Stochastic frontier analysis of state territory
public hospital systems

• Unlike DEA, SFA allow for measurement error, not
  just inefficiency
• The model estimated contains
• 1 Output (casemix-adjusted separations per
  jurisdiction)
• 3 Inputs (labour (FTE), real capital services,
  real medical supplies)
• Estimated in Stata using maximum likelihood
• Data from Australian Institute of Health
  Welfare Report on Government Service Provision
  Australian Bureau of Statistics
• All variables expressed per 1000 residents no
  adjustment for demographics
• Covers the period 1996-97 to 2004-05
• Alternative models
• Quality adjusted output (Casemix-adjusted
  separations adjusted by an index of
  life-expectancy at birth by state)
• Time invariant, Time variant

9
Public hospitals implied productivity gap by
state
10
Public hospitals implied productivity gap by
state
11
Summary

• Experimental results suggest that there could be
  scope for productivity improvement in Australian
  public hospital systems
• (Analysis suggest that this could be in the order
  of 10)
• Wide variation across jurisdictions
• However, caution needed
• Based on (sometimes dated) historical information
• Quality of data is less than ideal
• Do not isolate the effects of policy choices (eg
  achievement of equity goals) from efficiency and
  other influences
• Examination of the industry in situ, not forward
  looking do not fully take account of the
  potential for change
• Unable to control for all relevant institutional
  and operating factors

12
(No Transcript)