Improving the planning of the GP workforce in Australia: a simulation model incorporating work transitions, health need and service usage.

BACKGROUND In Australia, the approach to health workforce planning has been supply-led and resource-driven rather than need-based. The result has been cycles of shortages and oversupply. These approaches have tended to use age and sex projections as a measure of need or demand for health care. Less attention has been given to more complex aspects of the population, such as the increasing proportion of the ageing population and increasing levels of chronic diseases or changes in the mix of health care providers or their productivity levels. These are difficult measures to get right and so are often avoided. This study aims to develop a simulation model for planning the general practice workforce in South Australia that incorporates work transitions, health need and service usage. METHODS A simulation model was developed with two sub-models--a supply sub-model and a need sub-model. The supply sub-model comprised three components--training, supply and productivity--and the need sub-model described population size, health needs, service utilisation rates and productivity. A state transition cohort model is used to estimate the future supply of GPs, accounting for entries and exits from the workforce and changes in location and work status. In estimating the required number of GPs, the model used incidence and prevalence data, combined with age, gender and condition-specific utilisation rates. The model was run under alternative assumptions reflecting potential changes in need and utilisation rates over time. RESULTS The supply sub-model estimated the number of full-time equivalent (FTE) GP stock in SA for the period 2004-2011 and was similar to the observed data, although it had a tendency to overestimate the GP stock. The three scenarios presented for the demand sub-model resulted in different outcomes for the estimated required number of GPs. For scenario one, where utilisation rates in 2003 were assumed optimal, the model predicted fewer FTE GPs were required than was observed. In scenario 2, where utilisation rates in 2013 were assumed optimal, the model matched observed data, and in scenario 3, which assumed increasing age- and gender-specific needs over time, the model predicted more FTE GPs were required than was observed. CONCLUSIONS This study provides a robust methodology for determining supply and demand for one professional group at a state level. The supply sub-model was fitted to accurately represent workforce behaviours. In terms of demand, the scenario analysis showed variation in the estimations under different assumptions that demonstrates the value of monitoring population-based need over time. In the meantime, expert opinion might identify the most relevant scenario to be used in projecting workforce requirements.