CAN WE DESIGN A MARKET FOR COMPETITIVE HEALTH INSURANCE?
The empirical work on risk adjustment has most commonly focussed on predicting actual
expenditures. Predictive accuracy varies with the population considered (older people and those
with chronic illness have more predictable expenditures), the types of services (pharmaceuticals and
outpatient services), the time for which data are available (more recent data), and the length of time
over which expenditure is being predicted (longer periods). Predicting 100% of actual expenditure is
neither possible on current models, nor indeed desirable as discussed above; predictive accuracy
of current models is far less than 100% and more likely to be around 20% (van de Ven, Ellis 2000).
There has been limited research on risk adjustment in Australia; using NSW data and diagnostic
information, Donato and Richardson report predictive results of around 6% (Productivity Commission
2002).
There are strategies to lessen the impact of poorly performing risk adjustment models on selection
issues and insurance fund viability. These include various arrangements for reinsurance to cover
those individuals who do incur comparatively high expenditures. Such arrangements currently apply
in Australia for all privately insured patients over the age of 65 admitted to hospital, or younger
insured patients with more than 35 days in hospital (Department of Health and Ageing Private
Health Insurance Branch 2003). However, by applying reinsurance to the high use, high expenditure
group, the incentives for individual funds to better manage this group are muted. There is a trade-off
between appropriate incentives and fairness.
Risk adjustment can also be used as a research tool, and in allocating funds to population groups.
Pooled funding for population groups is discussed in the next section.
10