Abstract
This thesis addresses questions of how to incorporate quality of care, represented by
disutility-bearing effects such as mortality, morbidity and re-admission, in measuring
relative performance of public hospitals. Currently, case-mix funding and performance,
measured with costs per case-mix adjusted separation, hold hospitals accountable for
costs, but not effects, of care, creating economic incentives for quality of care minimising
cost per admission.
To allow an appropriate trade-off between the value and cost of quality of care a
correspondence is demonstrated between maximising net benefit and minimising costs
plus decision makers’ value of disutility events, where effects of care can be represented
by disutility events and hospitals face a common comparator. Applying this
correspondence to performance measurement, frontier methods specifying disutility
events as inputs are illustrated to have distinct advantages over output specifications,
allowing estimation of:
1. economic efficiency conditional on the value of avoiding disutility events.
2. technical, scale and congestion sources of net benefit efficiency;
3. best practice peers over potential decision makers’ value of quality; and
4. industry shadow price of avoiding disutility events.
The accountability this performance measurement framework provides for effects and
cost of quality of care are also illustrated as the basis for moving from case-mix funding
towards a funding mechanism based on maximising net benefit. Links to evidence-based medicine in health technology assessment are emphasised in illustrating application of
the correspondence to comparison of multiple strategies in the cost-disutility plane, where
radial properties as shown to provide distinct advantages over comparison in the cost-effectiveness plane.
The identified performance measurement and funding framework allows policy makers to
create economic incentives consistent with evidence-based medicine in practice, while
avoiding incentives for cream-skimming and cost-shifting. The linear nature of the net
benefit correspondence theorem allows simple inclusion of multiple effects of quality,
whether expressed as not meeting a standard, functional limitation or disutility directly.
In applying the net benefit correspondence theorem to hospitals a clinical activity level is
suggested, to allow correspondence conditions to be robustly satisfied in identification of
effects with decision analytic methods, adjustment for within DRG risk factors and data
linkage to effects beyond separation.