Measuring low-value care in New South Wales public hospitals

Abstract

My aim in this thesis is to develop and apply measures of low-value care (where expected benefits do not outweigh the potential harm) based on hospital administrative data in New South Wales (NSW), Australia. Measurement is a prerequisite for understanding the drivers of and exploring remedies for low-value care. I developed low-value care measures for 27 procedures. In 2016-17, between 4487 and 8986 hospital episodes involved low-value care, accounting for 10% to 20% of all episodes involving these 27 procedures. However, my results almost certainly underestimate low-value care for these procedures. The proportion of low-value care varied considerably between hospitals. Multilevel modelling analysis confirmed that most variation is between hospitals; Local Health District and residential area have little association with low-value care. None of the (limited) hospital variables available were associated with low-value care. Further exploration of inter-hospital variation will require more detailed hospital data, including attitudes of individual clinicians within hospitals. Low-value care is also a patient safety issue. I used 16 hospital-acquired complications (HACs) to examine some immediate adverse consequences of low-value care. Across seven procedures where recommended care would not normally involve hospital admission, HAC rates ranged from 0.1% to 15.0% of the low-value episodes. To the best of my knowledge, these patients only entered hospital for the low-value procedure; therefore, the HACs can be attributed to the low-value care. This project occurred in partnership with the NSW Ministry of Health. Two other states commissioned analyses using their own data, and a peak private health insurance industry body commissioned an analysis using private health fund data. State health departments are using these methods to provide feedback to clinicians and hospital managers and may in future incorporate the indicators into funding and performance agreements.