Preventable hospitalisations in Australia: understanding the impact of personal and health system factors using linked and longitudinal health data

Abstract

Preventable hospitalisations are used in Australia as a high-level indicator of health system performance, specifically the accessibility and quality of primary care. However, there are key gaps in understanding of how preventable hospitalisations relate to characteristics of patients and features of the health system, and surprisingly little evidence validating their use in Australia. In this thesis, new approaches to analysing longitudinal health data were applied to gain insights into the properties of this health performance indicator.

This thesis used linked questionnaire and longitudinal health data for a cohort of over 267,000 participants in the 45 and Up Study, Australia, containing detailed information on participants and their use of health services. Temporal patterns in use of primary care and other health services around preventable hospitalisation were explored using a visualisation of unit record health data. Predictors of preventable hospitalisation were identified using multilevel Poisson regression models, with variation partitioned between person- and geographic-levels. Through development of novel weighted-hospital service area networks , variation was further partitioned to the hospital-level.

Many patients admitted for preventable hospitalisation were found to have high levels of engagement with the health care system, both around the time of admission and compared to similar non-admitted patients. The supply of general practitioners explained only a small amount of geographic variation in preventable hospitalisation, while over one-third of variation was contributed by the sociodemographic and health characteristics of the population. Hospitals differed in their propensity to admit patients, with the greatest variability in smaller community hospitals, which account for a small proportion of admissions but contribute greatly to regional variation.

These findings show the preventable hospitalisation indicator in Australia should not be interpreted simply as a measure of the accessibility and quality of primary care. They suggest the most appropriate policy responses are long-term strategies to promote healthy living and targeted local interventions to efficiently manage the current burden of chronically ill patients. The findings demonstrate why caution should be used when adopting international health performance indicators, but also the benefits of using novel approaches to derive new information from linked and longitudinal data.